In this study, we reported a genetic screening method that evaluates the upper limit copy number of target genes. In this method, we used a gene with its native promoter as a unit to evaluate the upper limit copy number to inhibit cellular growth, so that we could quantitatively and directly compare the upper limits among various genes.
Principally, the gTOW experiment causes overexpression of target genes. In S. cerevisiae,
promoter system is the common way to overexpress target genes [27
]. We compared the growth inhibition in the gTOW experiment and GAL1
promoter expression system, but the data did not show complete positive correlations (C). The differences between the two experimental results are explained by the native expression level and the expression regulation of each target gene. If the native expression level of the target gene is high, the expression from GAL1
promoter does not cause so much “overexpression,” but increasing the copy number in the gTOW experiment does cause overexpression. For example, CLB3
did not show strong growth inhibition in the GAL
experiment but did in the gTOW experiment (C); probably because they have higher native expression levels [21
] (D). CLB6
are the opposite case [21
]. In addition, if there are transcriptional and translational regulations, such as periodic expression during cell cycle and feedback regulation within the target gene, the results of both experiments will be different. Recently, Sopko et al. reported a comprehensive overexpression analysis using the GAL1
]. Their results and the results obtained in our GAL1
experiment showed little accordance in growth inhibition (unpublished data). One of the reasons for this may be the existence of GST-tag in their experiment. In fact, we observed that the copy numbers in the gTOW experiments were perturbed by the commonly used epitope tags (unpublished data). Thus, when determining the quantitative effect of a target protein, a tagged protein is not preferred. Thus, the gTOW experiment enabled systematic evaluation of the upper limit of gene expression, about which little has been known hitherto.
The copy numbers of the 30 CDC genes determined in the gTOW experiment were very diverse, ranging from 1 to more than 100 (A). The data revealed some interesting properties of the cell division cycle in S. cerevisiae, which have been very difficult to clarify. Six out of the top seven genes with the lowest copy numbers constitute a subsystem that regulates B-type CDK activity, which was a core process in the cell cycle with very dynamic properties (A). We speculate that because the parameter range should be tuned up to be narrow in a system with very dynamic properties, the system should show high fragility against perturbations that change the quantity of parameters as a trade-off of the dynamics. In other words, the gTOW experiment was very effective to reveal dynamic subsystems that are fragile to changes in parameter.
To our surprise, CDC14
has a very low upper limit of less than two copies per haploid genome (i.e., just one extra copy other than the chromosomal one). Interestingly, this extreme low limit was almost consistent with that predicted by the computer model developed by Chen et al. (C), which could be explained by the 1:1 stoichiometric inhibition by Net1 [31
]. However, other stoichiometric partners, Esp1 and Pds1, did not show such extreme low limits, although the model also predicted very low limits (C). The discrepancy is probably explained by the fact that Esp1 needs to be recruited into the nucleus by Pds1 for its full function [33
], the regulation of which is not yet implemented into the model. If there is such regulatory mechanism in a system, the system should be rather robust even when the stoichiometric balance is perturbed. We do not yet understand why the CDC14
regulation evolved to be extremely fragile against the amount of change, but it might be a trade-off of some properties of the subsystem that CDC14
We used the data obtained in the gTOW experiment to evaluate a computer model. Generally, because intracellular biochemical parameters are very difficult to determine, it is difficult to evaluate the parameters in computer models. Since permissible ranges of parameters in a model are the integrative result of the network structure and parameters in the model, the parameter ranges are a very useful measure to evaluate the model's correctness and to suggest the direction to improve it [6
]. The models by Chen et al. showed much fragility relative to the gTOW experimental data. We suggest two major issues to be improved in the model: one is stoichiometric partners as mentioned above and the other is paralogous gene pairs. The model implemented only one of each paralogous gene pairs (i.e., CLB2
), but each of the paralogous gene pairs had very different copy numbers (i.e., upper limits) between them in the gTOW experiment (Figure S5
). The issue of how paralogous gene pairs are involved in cellular robustness is still being argued [34
] and it will be very interesting to test how robust the model becomes when the paralogous genes exist.
The gTOW method may also be used for genetic screening of positive growth regulators. Under mild copy number–increasing bias (i.e., uracil− condition), MIH1
had significantly higher copy numbers than the vector (B). Mih1 is known to dephosphorylate inhibitory phosphorylation of B-type CDK in Tyr-19 residue [23
]. In S. cerevisiae,
phosphorylation in Tyr-19 is involved in the morphological checkpoint [23
]. High copy number MIH1
may cause faster growth ignoring the morphological checkpoint. Interestingly, overexpression of Cdc25, the human homolog of Mih1, is known to be closely related to cancer development [37
], part of which might be related to factors in cancer development as in the case of MIH1
. Positive growth factors, which can be identified by gTOW experiments, under mild bias, potentially contain factors related to cancer development such as MIH1
We estimated the protein overexpression of about 19 Cdc proteins out of 30 tested in the gTOW experiment and confirmed that most of the Cdc proteins tested were overexpressed with a good agreement with their gene copy numbers (C). The overexpression of 12 Cdc proteins was appropriately measured with endogenous protein level (referred to as Class I proteins in this study, Figure S6
A and Table S5
). These will be good resources for further precise quantitative analysis, such as the protein expression in synchronized cells, or single cell variation in the gTOW experiment. Among Class I proteins, only Swi6 showed apparent discrepancy between protein overexpression level and gene copy number in the gTOW experiment (C); the mechanism how this occurs is also an interesting future issue. Current epitope tags used for protein detection were not preferable for the perturbation analysis such as gTOW (unpublished data), but specific antibodies were not comprehensive or qualitative enough; we thus need more comprehensive and qualitative technology to detect proteins with lowest perturbations.
In conclusion, using the gTOW method, we obtained upper limit gene copy numbers, about which little has been known at a system-wide level. This sort of quantitative data represent how intracellular parameters are set up in a certain biological system, and thus represent the robustness and fragility of the system against internal perturbations, which have been very difficult to assess with experimental data. In addition, as shown here, the data can be used to evaluate computer models and to improve them. The gTOW method can also be applied to genome-wide analysis of upper limit gene copy numbers, other than those of CDC genes, as well as to profiling of quantitative genetic interactions in mutant strains.