1. Presentation by:
Kyle Borge, David Byon, & Jim Hall
Herpesviral Protein Networks and Their Interaction with the Human Proteome
Jim
reconstruction of a Herpes Virus capsid
Presentation by:
Kyle Borge, David Byon, & Jim Hall

2. Introduction to the Herpesvirus
Large double-stranded DNA genomes
Eight different strains
Causes diseases ranging from cold sores to shingles
Vaccine available for Varicella-Zoster Virus (VZV)
Little known about protein interactions
Jim Little is known about herpesviral protein interactions
Herpesviruses are divided into three subfamilies; α - alphaherpesvirus (three viruses including VZV), β - betaherpesvirus (two viruses) & γ - gammaherpesvirus (two viruses; KSHV and Epstein-Barr Virus(EBV))

3. Types of Herpesviruses Investigated
Kaposi’s Sarcoma-associated Herpesvirus (KSHV)
In the gamma (γ) herpes virus phylogenetic class
Causes cancerous tumors
Mostly associated with HIV patients
Sequenced in 1996
Genome is roughly 165 kbs
89 open reading frames (ORFs)
113 ORFs used in experiment (included 15 cytoplasmic and 5 external domains derived from transmembrane proteins)
Varicella-Zoster Virus (VZV) , in the alpha (α) herpesvirus phylogenetic class
Causes chicken pox in children and shingles in adults
Sequenced in 1986
Genome is roughly 125 kbs
69 open reading frames (ORFs)
96 ORFs used in experiment (Included 13 cytoplasmic
and 10 external domains derived from transmembrane proteins)
Jim

4. Methods of Investigating Protein Protein Interactions (PPI)
Many Methods
The Y2H technique is one of the top techniques for detecting protein-protein interactions
This article used Y2H to investigate protein-protein interactions
Kyle

5. Y2H Advantages http://www.dnatube.com/video/993/Plasmid-Cloning
Relatively simple (automated)
Quick
Inexpensive
Only need the sequenced genome (or sequence of interest)
Scalable, its possible to screen for interactions among many proteins creating a more high-throughput screen (ex. viral genome)
Protein/polypeptides can be from various sources; eukaryotes, prokaryotes, viruses and even artificial sequences…allows the comparison of interactomes w/in and between different species…in this paper, eukaryote (human) interactome vs. viral interactome
Kyle

6. Y2H Limitations http://www.dnatube.com/video/993/Plasmid-Cloning
The Y2H system cant analyze some classes of proteins
Transmembrane proteins, specifically their hydrophobic regions which may prevent the protein from reaching the nucleus
Transcriptional activators; may activate transcription w/out any interaction
False-negatives
Y2H screen fails to detect a protein-protein interactions
False-positives
Y2H screen produces a positive result (characterized by reporter gene activity) where no protein-protein interaction took place
Ex. bait proteins activate, transcribing the reporter gene, w/out the binding of the AD (bait proteins act as transcriptional activators)

7. Yeast’s GAL4 transcriptional activator
GAL4 transcriptional activator which splits into two separate fragments; a binding domain (BD) and an activating domain (AD)
Kyle

8. Y2H Method ORFs selected from published sequences
Amplified by nested PCR
Made primer sets of ends of ORFs
Y2H bait and prey vectors
Vectors transformed into Y187 and AH109 haploid yeast cells creating pools; a bait pool and a prey pool
Bait and prey mated in quadruplicates
Positive diploid yeasts are selected
- Kyle
- Viral genes selected from published sequences
- pDEST-GADT7 and pDEST-GBKT7 into bait and prey vectors -

9. Open Reading Frames (ORFs)
Every ORFs of both KSHV & VZV were cloned & ligated into both a bait and prey GAL4 vector
Bait
protein of interest
the protein is fused to the yeast Gal4 DNA-binding domain (DBD)
Prey
a protein/ORF fused to the Gal4 transcriptional activation domain (AD)
interacting protein
Physical interaction between the bait and prey brings the DNA-BD and an AD of Gal4 together, thus re-creating a transcriptionally active Gal4 hybrid
Gal4 activity can be assayed by the expression of reporter genes and selectable markers
Kyle

10. (1-2) ORFs cloned into vectors via Nested PCR
KSHV
113 full-length and partial ORFs
including 15 cytoplasmic and 5 external domains derived from transmembrane proteins
VZV
96 full-length and partial ORFs
including 13 cytoplasmic and 10 external domains derived from transmembrane proteins
- Kyle
- Turns blue to indicate interaction between viral proteins

11. Yeast-Two-Hybrid Bait pool:
Prey pool: (target)
Each individual ORF sequence is cloned into the ‘prey’ vector (down stream of the GAL4 AD gene) and is essentially fused to the GAL4 AD gene
Ampr for selection
Hemagglutinin
Bait pool:
Each individual ORF sequence is also cloned into the ‘bait’ vector (down stream of the GAL4 DBD gene) and is essentially fused to the GAL4 DBD gene
vector conveys Kanr for selection
Amp & Kan genes of these plasmids ‘give/promote’ Ampr & Kanr resistance, individually…and later on, they will permit selection for diploid cells (from the two haploids mating)
Both the KSHV & VZV genomes had been previously sequenced which enabled researchers to identify the ORFs as they have a couple different characteristic sequences organisms with dsDNA…obtained from other cells
ORFeomes for both viruses were amplified via nested PCR (giving better assurance for ORF sequence specificity)
PCR product were cloned by BS clonase into the gent

12. Yeast-Two-Hybrid Background
…in a diploid cell.
Amp & Kan genes of these plasmids ‘give/promote’ Ampr & Kanr resistance, individually…and later on, they will permit selection for diploid cells (from the two haploids mating)
Both the KSHV & VZV genomes had been previously sequenced which enabled researchers to identify the ORFs as they have a couple different characteristic sequences organisms with dsDNA…obtained from other cells
ORFeomes for both viruses were amplified via nested PCR (giving better assurance for ORF sequence specificity)
PCR product were cloned by BS clonase into the gent

13. Viral Protein Interactions in KSHV
12,000 Viral Protein Interactions tested
Identified 123 nonredundant interacting protein pairs
118/123 were novel
7/123 were previously reported
Screen captures 5/7 (71%) of previously reported interactions
50% of Y2H interactions confirmed by coimmunoprecipitation (CoIP)
DAVID
Of the 123 unique interactions detected in KSHV, 4.1% (5/123) were previously reported (additional 2 were known by personal communication) and 95.9% (118/123) novel, of which 50% (59/118) could be confirmed by CoIP. 71.4% (5/7) of the previously known KSHV interactions were captured by the Y2H screen.

14. Viral Protein Interactions in KSHV
David
Fig. S1
The red boxes indicate the mirror axis – same protein tested for interaction
The yellow boxes are self-activating baits
The black boxes are positive for interaction between the two proteins
The number of interactions (black boxes) are added up on the axis of the red box
Two clusters of protein interactions can be noticed here and here. Take note there are very few clusterings.

15. Previously Reported Protein Interactions of KSHV
David
Table S2
Previously reported data
The columns from left to right depict the two proteins tested for interaction, the protein type, where it was referenced, and the results on Y2H and CoIP.

16. Coimmunoprecipitation
JIM
As good a method Y2H is in detecting protein-protein interactions, it still has limitation, and due to these limitations (false-negatives, false-positives, ect.) involved with Y2H assay, protein interactions detected by the Y2H analysis need to be varified via other qualitative experiments, such as CoIP; and in fact, of all the KSHV protein-protein interactions detected by Y2H (as well as by orthologous proteins in HSV-1, VZV, CMV, and EBV), 48% were verified by a beta-galactosidase assay adn 48% were verified by a CoIP. And the same was done for VZV protein-interactions, predicted KSHV-human protein interactions, ect.
CoIP was determined by...
- bait are myc tagged, prey are ha tagged
- potential interacters are first precipitated with one antibody, then immuno-labeled with the other.
- only co-precipitaed proteins will bind the immuno label.

17. Verification of Predicted Interactions in other Herpesvirus Species
Jim - since only 50% or Y2H interactions could be confirmed by co-immunoprecipitation, the researchers looked to other viruses with orthologs that could be confirmed. n.d. means “not done”
Table S3

18. Correlation Between Viral Protein Interaction and Expression Profile
Figure S3: Jim – table on the left was created from someone else’s work looking at expression profile for 81 ORF’s, yellow show that they have similar expression profile, where red ones are more distinct. The small circles in this graph show interacting protein pairs. From this data expression profile correlation was calculated.
In the graph on the right, this compares the correlation between AEC and clustering coefficient. Size of circle represents number of interacting partners, color represents functional class, If interactions are static such as in a large complex proteins are more likely to interact with each other (hence higher C), and to be expressed at the same time as their interacting partners, hence higher AEC. On the other hand if the interactions take place at a different time/place, both C and AEC should be lower. “Party vs. Date” hubs.
If i
Fig. S3
Average expression correlation [AEC]was calculated
For random pairs of ORFs:
For interacting pairs of ORFs: 0.839
Correlation between AEC and clustering coefficient
Used to propose static or dynamic interaction for viral hubs

19. Protein Interaction Networks
David
Since the researchers now had protein interaction data from KSHV and VZV and also their interactions with orthologous proteins in other herpesviruses, they further investigated herpesvirus protein interactions by constructing networks of the viral proteins and then predicted their interactions with the human protein network.
The figure shown here could represent a typical protein network. The circles represent nodes, the lines edges,

20. Network Terminology Node – represents a protein
Edge – represents interaction between two nodes
Average (node) degree – the average number of neighbors or connections that any given node has
Power coefficient (g) – derived from an approximate power law degree distribution plotted on a bilogarithmic scale and fitted by linear regression
P value - (significance under linear regression) as fitted by a power-law degree distribution (‘‘scale-free’’ property)
Characteristic path length – the distance between two nodes
Diameter (d) - describes the interconnectedness of a network; defined as the average length of the shortest paths between any two nodes in the network
Clustering coefficient – A value given to depict the number of fold enrichment over comparable random networks (‘‘small-world’’ property)
Small world property/network – Any network that has characteristics of a relatively short path and dense cluster (high cluster coefficient)
-So the power coefficient is used in this paper to compare the herpesviral protein networks to both yeast and human protein networks. The power coefficient is derived from plot of the log for the node degree vs. the log of the probability or frequency of node degree. The plot depicts a noticeable difference between the virus and human/yeast protein interactive network. It is more likely for the viruses to have nodes with higher degree than humans and yeasts.
-The diameter characterizes the ability of two nodes to communicate with each other: the smaller d is, the shorter is the expected path between them
Networks with a very large number of nodes can have quite a small diameter

21. Topology of KSHV and VZV Interaction Network
KSHV protein interaction network
David
VZV protein interaction network

22. Comparison of Protein Interaction Networks
David
Table 1
This table shows the comparison of network parameters of eight different organisms and viruses using a power law distribution. Node indicates protein; edges are number of interaction between nodes; average degree is the average number of interactions per node; the power coefficient and P value determines the relative stability or resistance of a particular organism or virus under a linear regression; characteristic path length is the relative distance between nodes; the diameter is the maximum distance between any two nodes; the clustering coefficient is a value depicting the

23. Power Law Distribution Comparison
David - Figure 1B
This figure is a comparison of the approximate power law distribution of two herpesviruses, yeast and human protein interaction networks. The x-axis is the log of k which is the node degrees. The y-axis is the relative frequency (i.e. probability). A linear regression was fit to the datasets. Can everyone see the difference between the viral and yeast/human datasets? (They say yeah I do see a difference.) This shows that organization of the interaction networks of viruses are significantly different than from yeast and human.

24. Removal of Nodes in KSHV Network
David
Fig. 1c&d
These figures represents the removal of nodes and its overall effect on path length and network size of yeast and KSHV.
The figure on the right has an x-axis with percentage values of hubs removed. Hubs are nodes with many interacting partners. The y-axis represents network path length. As the percentage of nodes removed increases there is a significant difference between the trail of KSHV and yeast. There is a noticeable fluctuation in path length of the yeast network. In the graph to the right a very similar picture unfolds. The x-axis still remains the same as the left graph as its plotted against a relative network size. The yeast show steep decline in its network size as more hubs are removed relative to KSHV. Both graphs depict the yeast network to become more unstable as more hubs are removed. This suggest KSHV to be more resilient than their yeast counterpart.
Path length – the distance between two nodes
This path length represented here is the average path length between any two nodes. Removing highly connected nodes from yeast when compared with KSHV greatly reduces network size since they are small
Figure 1C. Depicts as the percentage of hubs removed increases yeasts have

25. Protein Interaction KSHV & Sequence Conservation to EBV
Jim
Fig. s7b

26. Correlation Between Functional and Phylogenetic Herpesviral Classes
Fig S8a - Jim
KHSV ORF’s were partitioned into five functional classes. Then they were further partitioned into whether they were KSHV specific or had orthologs n

27. Viral protein interactions between functional classes
S8B - Kyle
Size = the number of proteins within the functional class
Interactions with = the number and functional class of the interactors
Interestingly enough, more then 70% of all protein protein interactions occurred between viral proteins belonging to different functional classes…for instance, as can be seen from the graph, the number of proteins involved in replication was approximated to equal a size of 10. However, the proteins that interact to influence replication are very diverse (are proteins that represent many different functional classes);
In general, interacting proteins are not more likely to belong to the same functional class than random pairs of proteins, suggesting that the majority of viral proteins either have multiple, previously unknown functions, or are assigned incorrectly.
significantly increased numbers of intra-class interactions were only identified between proteins belonging to the gene regulation and unknown functional classes (statistics see Table S7)….
the most significant correlation between, proteins that belong to the same functional group interacting (PPI) to and ‘effecting’ that specific function was seen in regulation and unknown… For instance, gene regulation. The size (# proteins within functional group) is almost similar to the # gene regulation proteins that actually interact to influence/effect gene regulation (or a gene regulation event)
?Intra-class interactions - protein of certain functional group interact to influence that actual function?
Whereas interactions between proteins involved in host interaction were most highly suppressed (as expected)
interactions between proteins involved in both host interaction and replication were significantly increased
Interactions between proteins with unknown function are significantly overrepresented (among each cellular function), implying that they are within the same functional group (other than host interaction) and parts of identical complexes or processes

28. Viral Protein Interactions Between Phylogenetic Classes
S8c - kyle
FYI…The β and αγ subclasses, which contain only 1 member each, do not provide additional information and are thus not shown
Herpesviruses are divided into three subfamilies; α - alphaherpesvirus (three viruses including VZV), β - betaherpesvirus (two viruses) & γ - gammaherpesvirus (two viruses; KSHV and Epstein-Barr Virus(EBV))
Phylogenetic classes were partitioned depending on whether they are specific for KSHV
OR possess orthologs in the γ (γ, alphaherpesvirus subfamilies), β and γ (βγ, alphaherpesvirus & betaherpesvirus subfamilies) or α, β and γ (αβγ, alphaherpesvirus, betaherpesvirus & gammaherpesvirus subfamilies) subfamilies

29. View of the Human-Herpesviral Networks
Varicella-Zoster Virus
Kaposi Sarcoma-associated Herpesvirus
David
Figure S10
The viral proteins are depicted by red nodes and the cellular interacting human proteins or target proteins are depicted by the blue nodes.
Predicted viral interactions are shown with red edges (or lines connecting the nodes), cellular interactions are blue edges and viral-human node interactions with green nodes.
Both VZV and KSHV appear to show very similar predicted network patterns which adds some fidelity to their approach.

30. Power Coefficient of KSHV-Human Network
David
Figure S9
This is a power graph depicting viral-host network level vs. the power coefficient. The red line is the power coefficient for KSHV protein network and the blue line is the power coefficient of the predicted human protein network. The green line starts at level 1 which is where the viral protein interacts with the target human protein. It steadily increases until level 6 where the power coefficient is very close to the protein coefficient of the human network.
So why does

31. Interplay between KSHV and Human Network
David
Figure 2A.
This figure represents the global topology of the integrated KSHV network into the interacting human network. Red nodes and edges depict viral proteins and interactions them; blue nodes and lines represent level 1 or 2 interacting cellular proteins and interactions between them; gray nodes are cellular proteins greater than level 2 (nodes that are indirectly affected by viral nodes). There are 10,636 edges and 3,169 nodes which complements

32. Viral Host Network / Random Network Comparison
Jim Figure 2C: Random
Combined Viral /Host network was compared to 1000 random networks, that were generated by rewiring fixed virus interactors to swapped cellular proteins with the same degree as actual target. Random networks are blue circles, KSHV/Human network is a red triangle. What is noted is that only 65 random networks have a higher Power coefficient

33. Conclusions
Virus and host interactomes possess distinct network topologies
Integration of viral and host protein network may lead to better understanding of viral pathogenicity
Future interactome data from other viruses may improve understanding of functions of viral proteins and their phylogeny
Understanding networks may help to develop future therapies
David