Although DNA replication and repair occur with impressive fidelity, the sheer amount of information that must be copied ensures that every time a genome is replicated, heritable mistakes are made. Over time, these errors and larger scale changes in chromosomes such as transposon movement, repeat expansion and contraction, and chromosome rearrangements have led to significant genetic diversity in populations. These genetic alterations have resulted in complex heterozygosity at many loci that are re-assorted every generation, resulting in a moving target of complex genetic interactions whose influence on phenotype is difficult to ascertain. Many, but not all, mutations are silent but it has been estimated from deep sequencing of 179 human individuals, that we all inherit 250–300 loss-of-functional alleles. Cancer cells represent perhaps the most extreme case of genetic diversity, as they frequently lose whole chromosomes or whole sections of chromosomes, resulting in the potential for thousands of complex haploinsufficient interactions. In our work we have been attempting to model CHI interactions in a simple eukaryote and specifically using the single essential actin gene, which plays critical roles in a large number of cellular functions. Our results confirm that there is a tremendous potential for deleterious CHI interactions in the genome. However, actin is so centrally important, we cannot say that all genes will be so genetically promiscuous.
Genetic interaction analysis has been used extensively in yeast to attempt to identify functionally related genes, the well supported assumption of course being that a binary gene interaction suggests relatedness. Such screens have led to a fairly anecdotal lore about what a particular type of interaction means, e.g. suppressor screens tend to identify gene products that physically interact or at least operate in the same pathway. The tremendous amount of synthetic lethal data generated by the Boone laboratory has pushed the ability to formalize rules about synthetic lethal interactions and how patterns of such interactions can be accurately used to determine function in same or parallel pathways 
. Although very few CHI screens have been performed at this time, we expect that many of the lessons learned from synthetic lethality will hold true.
It is not surprising that a null allele for actin would have such a large number of CHI interactions; the actin cytoskeleton is centrally important to many cellular functions. Further, actin can exist in many highly dynamic forms that are regulated in complex spatio-temporal ways by a large number of associated proteins. Interpretation of any single CHI interaction will need to take into account these complexities. One thing that is clear is that CHI screens, like synthetic lethal screens, identify groups of functionally related genes and, with respect to actin, we can conclude that some very important cellular processes are hyper-sensitive to reductions in actin expression.
Since actin itself is haploinsufficient, we may have expected our screen to be overwhelmingly biased toward other genes that also display simple haploinsufficiency. However, this does not appear to be the case, thus bolstering the argument that, like other genetic interactions, a CHI interaction reflects functional specificity. Another possible explanation for CHI interactions could be cumulative reduced biosynthetic capacity 
, but our original CHI screen against the non-essential genes was not overly enriched for genes involved in biosynthesis, with the exception of the ribosome. The CHI screen against the essential genes presented here did identify a number of genes whose products are components of the proteasome and the core transcriptional apparatus, but we could not detect significant changes in the levels of actin, or the actin binding proteins Aip1p and cofilin in strains hemizygous for RPN5
, or RPS5
(data not shown) suggesting these interactions may reflect true functional connections.
A common interpretation of synthetic lethal interactions between null alleles is that the two genes operate in parallel pathways that impinge upon an essential function 
. A fundamental difference of a CHI interaction is that since activity from both genes is reduced by at most 50% and not 100%, the pathways the genes function in are not blocked entirely. Therefore, we might expect that a CHI interaction may frequently be indicative that the two genes/gene products function within the same pathway or structure; that pathway function is compromised by loss of flux due to constrictions at two steps. Given the striking enrichment for proteasome genes and the novelty of a proteasome-actin connection, we chose to investigate a potential functional connection between the actin cytoskeleton and the proteasome.
Analysis of Ts− proteasome mutants showed that many, but not all, have severe defects in actin organization. In some cases these defects are so severe that this aspect of the phenotype alone might be expected to cause cell death. The data suggest, from two perspectives that the actin phenotypes cannot be purely explained by losses in proteasome proteolytic activity. First, treatment of sensitized cells with the proteasome proteolytic activity inhibitor MG132 resulted in very distinct defects in cell polarity and actin polarization that differs from that observed in Ts− proteasome mutants. Second, several Ts− proteasome mutants that cease growth entirely at 37°C arrest without any actin organization defects. We hypothesize that the phenotypic diversity observed in the proteasome mutants likely reflects differential effects of the alleles on multiple proteasome activities. For example, loss of proteolytic activity may indirectly affect actin and cell polarity. However, we hypothesize there is a second function that is evident when there are defects in a direct interaction between actin and proteasomes.
This hypothesis is supported by actin filament pelleting assays suggesting a direct physical interaction between the proteasome and actin filaments, although we cannot rule out possible non-specific interactions between two very large complexes. However, we failed to see pelleting of a BSA control with actin filaments indicating that actin is not merely “sticky” under these assay conditions (data not shown). Nonetheless, the data suggest that there are F-actin binding sites on both the 19S RP and the 20S CP. However, since mass-spectrometry identified 19S proteins in the 20S preparations, we cannot preclude the possibility that the 19S RP was bridging an interaction between F-actin and the 20S core in our assays. In either case, either double-capped 26S proteasome, or the 20S CP could be a bivalent actin filament cross-linking protein. This model agrees well with EM images showing a strikingly ladder-like cross-linking of rabbit muscle actin filaments by rabbit reticulocyte proteasomes 
. These data suggest that the actin-proteasome interaction is conserved, which is supported by co-localization of proteasomes with the actin cytoskeleton in many different cell types including Xenopus
, epithelial cells and fibroblasts 
, myoblasts 
, and in the sarcomeres of muscle cells 
One curious aspect of our proteasome data is that not all proteasome genes are CHI with actin and not all Ts−
alleles in proteasome genes cause actin organization defects. In our previous work on CHI interactions between actin and the non-essential gene knock-outs 
, we uncovered functional divergence between ribosomal paralog genes for actin-related defects. It has been shown that proteasomes with an alternative composition can be assembled in certain genetic backgrounds 
. However, we do not believe that our data suggest the existence of a proteasome of alternative composition that has actin-specific functions. Instead, given that proteasome assembly has been shown to occur in a series of ordered steps 
, , we hypothesize that subunit limitation in hemizygous strains or mutation of certain proteasome subunits leads to the accumulation of assembly intermediates that selectively affect actin-specific functions of the proteasome. In regards to the phenotypic differences between Ts−
mutants of proteasome component genes, this likely reflects differences in the defects caused by the alleles. For example, some Ts−
mutants may affect function without affecting structure while others may affect both function and structure. Given our results with MG132 and actin organization, we hypothesize that proteasome structure is required for proper actin organization but catalytic activity may not be required for proper actin organization.
In summary, our results clearly prove the utility of CHI screening for making novel connections between cellular subsystems, for predicting gene function, and for mapping the landscape of binary gene interactions that are likely to be relevant to system collapse in more complex organisms including multigenic influences in human genetic disorders. In particular, extensive CHI interactions between actin and proteasome genes predicted a novel role for the proteasome in affecting actin cytoskeleton organization.