This page was produced as an assignment for an undergraduate course at Davidson College.
Physical and in silico approaches identify DNA-PK in a Tax DNA-damage response interactome: Paper Summary
Human T-Cell Leukemia Virus Type 1 (HTLV-1) causes Adult T-Cell Leukemia, presumably through the viral protein product Tax interacting with human cells. Tax was discovered in 1982 and many papers have suggested its exact functionality, raising much controversy. Several methods were used in previous papers to find cellular proteins that bind with Tax, including the yeast two-hybrid approach, immunoprecipitation, western analysis, 2D gel separation, and MALDI-MS. After conducting a literature review for confirmed Tax-binding proteins, the authors found 67 proteins of interest. The authors limited these proteins to Tax-binding proteins that could possibly be involved with DNA repair response. Four proteins of interest were identified: Rad51, TOP1, Chk2, and 53BP1. These four proteins were the basis from which Ramadan and colleagues began constructing an in silico Tax interactome.
The researchers then wanted to make a physical Tax interactome. They began by constructing a fully functional S-Tax-GFP fusion protein (S-tags and His6 on the amino end of the protein, and GFP fused on to the carboxyl end). The GFP was used to view the protein’s localization when it bound to cellular proteins, and the S-tags were used to purify Tax and its bound proteins with S-agarose beads (an immunoprecipitation procedure). The procedure was optimized to identify only those proteins that were specific to binding Tax. Then, the Tax-cellular protein complexes were trypsinized and underwent LC-MS/MS analysis (this is a way to identify proteins in a liquid using liquid chromatography and mass spectrometry). This process was done three times, and 86% of all binding proteins identified bound on all three runs. Twenty-five proteins bound to the control S-GFP fusion. Of the proteins that bound to the S-Tax-GFP fusion on all three runs, the researchers wanted to quantify which interactions were the strongest. They used data from mass spectrometry to determine the peptide sequences, and calculated a ranked list of bound proteins. The top five proteins were DNA-PK, Vimentin, Gamma interferon-inducible protein, PARP, and H2A.1 (a histone). Identifying DNA-PK, a DNA-dependant Protein Kinase, as a Tax-binding protein was a novel finding.
After assimilating a database of Tax interactions and a physical interactome, the researchers wanted to define the first neighbor interactions of known Tax-binding proteins. They began with the four database identified proteins (Rad51, TOP1, Chk2, and 53BP1) and used HRPD to find their first neighbor interacting proteins. This network consisted of 50 proteins involved in 112 interactions, graphically depicted in Figure 1 and referred to as G1. These 50 proteins include a DNA-PKcs (the catalytic subunit of DNA-PK, two regulatory subunits are needed to complete the DNA-dependant protein kinase) in a large core of interacting proteins, indicating an important role for this group in the Tax interactome.
Figure 1. The first neighbor interactome of the four proteins identified by the database as binding to Tax with roles in DNA repair. Proteins are represented by circles, and interactions are represented by lines. The four proteins are highlighted in yellow. The DNA-PKcs (PRKDC) is also highlighted. Some interactions around the periphery are unclear because of the need to have lines go underneath the protein circles.
The researchers then wanted to look at the interactome if the original four proteins were removed. Pulling out Rad51, TOP1, Chk2, and 53BP1 produces the interactome pictured in Figure 2 below.
Figure 2. The largest interactome of first neighbors of Rad51, TOP1, Chk2, and 53BP1, without including the actual four proteins. PRKDC is the protein kinase catalytic subunit, and it is highlighted in yellow. PRKDC is one of the top six proteins when considering length from one protein to all others and the amount of other proteins it interacts with. This figure shows that DNA-PKcs' centrality to DNA-repair mechanisms is independent of the four artificially chosen proteins.
Next, the authors wanted to show what the G1 interactome would look like if it only included proteins known to be primarily involved with DNA repair. The made an interactome composed of 26 proteins and 42 interactions as shown in Figure 3.
Figure 3.This is the interactome of first neighbors of Rad51, TOP1, Chk2, and 53BP1, with proteins that do not have a role in DNA repair removed. This interactome is referred to as G1* by the authors. DNA-PKcs is involved in the fourth most interactions of any other protein on this graph, and it has the ninth shortest overall distance to all other proteins. This indicates, once again, that DNA-PKcs are very central to this interactome.
The scientists then extended their consideration to second neighbor proteins - proteins that interact with the proteins that interact with Rad51, TOP1, Chk2, and 53BP1 - of the C1 graph. This network, referred to as G2, has 667 proteins and 3827 interactions. They restricted this network to genes involved in DNA repair to make G2*, which consists of 114 proteins. They then clustered the interactions to try to determine which proteins were the most central to the G2* network, only considering proteins with 3 or more interactions. They found five clusters, and idenitifed three proteins that bridge the five clusters together - one of which is DNA-PKcs.
Figure 4.This is the interactome of second neighbors of Rad51, TOP1, Chk2, and 53BP1, with proteins that do not have a role in DNA repair removed. The proteins were clustered based on interactions to simplify the graph. Three bridge proteins connect the five different clusters of proteins, and these bridge proteins - PRKDC (DNA-PKcs), PCDNA, and TP53 - are central to the interactome. Proteins in a cluster tend to have similar response mechanisms. For example, Cluster One contains for proteins that fix interstrand crosslinks in DNA. Cluster Two contains genes that stop the cell cycle in response to genotoxic stresses. Tax may affect all of these systems' abilities to respond to DNA damage.
The scientists then wanted to verify DNA-PKcs ability to bind to Tax. They performed an affinity pull-down assay (which are similar to immunoprecipitation) where Tax was used as the "bait" to find proteins in a cell that bound to it. The extracted proteins were separated on SDS-Page and subjected to immunoblotting (Western Blotting) with anti-DNA-PKcs.
Figure 5. This is the immunoblotting results for S-GFP (a control) and S-Tax pull-down assays. On Lanes 1 and 3, the researchers ran all the extracted proteins from the cells they analyzed before the pull-down assay. In lanes 2 and 4, they ran all the protein they extracted that was pulled down by S-GFP or by S-Tax. The first and third horizontal gel image shows where Tax and GFP should be on the gel. The middle gel image shows the immunoblot with DNA-PKcs. DNA-PKcs are found in the total protein extract from both the S-GFP and S-Tax affinity tests, which is expected, but after the affinity tests, DNA-PKcs are only found on the S-Tax extracted proteins. This shows that DNA-PKcs were extracted by affinity binding to Tax.
Overall, the scientists identified proteins that bound to Tax. They limited their analysis to proteins that were involved with DNA repair, as Tax was previously shown to interrupt DNA repair mechanisms. After performing analysis on these proteins, they found that DNA-PKcs was an important, novel protein associated with DNA repair mechanisms and Tax. Using in silico models as well as proteomic analysis will lead to a better understanding of Tax's role in HTLV-1 and DNA repair.
I think that the researchers sufficiently showed that DNA-PK was an important protein in the Tax interactome, but they did not really tell us why that was important - do they expect that some sort of HTLV-1 treatment will be a result of this research? Also, according to their paper, I though they made two separate Tax interactomes: one based off of previous literature, and one based off of their findings. However, I did not really understand how the two different interactomes worked together and had trouble discerning at times throughout their paper which interactome they were referring to. It seemed as though they did not really use their physical Tax interactome map. I wish they had more clarification on how their physical Tax interactome map was used in conjunction with the in silico map created from previous literature about Tax.
Works Cited
Ramadan, E, et al. (2008). "Physical and in silico approaches identify DNA-PK in a Tax DNA-damage response interactome". Retrovirology 5:92.