Evaluation of Upper Limit Copy Numbers of CDC Genes by the gTOW Method

As a test case, we analyzed the 30 cell division cycle–related (CDC) genes listed in Table 1 with this method. These genes are involved in the regulation of cyclin-dependent kinase (CDK) activity through the cell cycle [19,23]. We cloned each CDC gene with its native regulatory DNA elements, such as promoter and terminator, so that the regulation of the gene expression can be comparable to the chromosomal copy. Because these elements have not been fully determined so far, we cloned each gene with upstream and downstream DNA sequences up to the neighboring open reading frames (ORFs) into the pTOW-target. We measured the maximum growth rate of yeast BY4741 cells [24] transformed with each of these pTOW-CDCgene plasmids in the leucine− condition. We then determined the plasmid copy number (i.e., the gene copy number on the plasmid) within the cells cultured for 50 h as follows: The total DNA extracted from the cells was tested by two quantitative PCR with two primer pairs that amplify fragments of LEU2 on the plasmid and LEU3 on the chromosome, respectively. Then the ratio of the amount of LEU2 to that of LEU3 was calculated. Thus, the plasmid copy number is per haploid genome, and the actual copy number of the target gene is the plasmid copy number plus one because there is one extra copy on the chromosome.

Table 1

Cell Division Cycle–Related (CDC) Genes and Their Plasmids Used in This Study

Open in a separate window

Figure 1B is a scatter plot showing the relationship between the growth rate and the copy number of the 30 CDC genes determined in the gTOW experiment. The copy numbers and growth rates of genes with copy numbers of less than 60 showed a linear correlation. This indicates that the obtained copy numbers of these genes were determined by gTOW as shown in Figure 1A (a hypothetical mechanism for this linearity is shown in Figure S1). Genes with more than 60, but less than 80 plasmid copies, did not show obvious growth retardation, but the copy numbers are probably determined by gTOW as well (see below). Figure 2A shows the plasmid copy numbers of 30 CDC genes obtained in the gTOW experiment. They were diverse, from less than one to more than 100. The plasmid vector without a target gene showed about 160 copies per haploid genome. Genes with a copy number close to the vector probably do not reach the upper limit. Some cells with genes of low copy numbers showed abnormal morphologies such as cell elongation. Those morphologies were similar to the ones which have been observed when CDC genes are overexpressed (unpublished data), supporting the idea that the proteins encoded on the target genes were overexpressed and determined the copy number in the gTOW experiment.

Open in a separate window

Figure 2

Upper Limit Copy Numbers of CDC Genes

Copy number of the plasmid with each of the 30 CDC genes was determined in the gTOW experiment in leucine− (A) and uracil− (B) conditions. *The copy numbers of CLB3 and CLB2 were determined in the conditions with 40 μg/ml and 20 μg/ml leucine, respectively. The data used in the graphs are listed in Table S2.

The cells with very low copy plasmid numbers (<35) hardly grew in the leucine− condition, probably because the number of leu2d is too low to support the leu2 deletion. When they were cultivated for a very long time, revertant cells were sometimes observed. In this case, we further evaluated the upper limit by adding a low concentration of leucine into the growth medium in order to reduce the bias toward increasing the plasmid copy number (Figure 3). Along with the increase of leucine concentration in the medium, growth of the cells with the vector alone progressively increased, and the plasmid copy number in the cells decreased from about 160 to 40 copies per haploid genome (vector in Figure 3). We tested some genes with low plasmid copy numbers in the leucine− condition. Genes such as CLB2, CLB3, CLB5, MCM1, SIC1, and SWE1 showed a dramatic switch-like decrease of growth below certain leucine concentrations (Figure 3). This switch-like decrease may be due to the nature of the regulation of these genes with positive feedbacks (see below). In these cases, we may have to regard the copy number in the leucine concentration just before the dramatic growth retardation as the upper limit.

Open in a separate window

Figure 3

Determination of Upper Limit Gene Copy Number with Leucine Supplementation

Growth and copy number of plasmids in the gTOW experiment with various leucine concentrations are shown. The data used in the graphs are listed in Table S2.

As noted above, the plasmid copy numbers in the uracil− (leucine+) condition also seem to be determined by weak gTOW. In this condition, growth retardation due to maintaining the plasmid was not observed except CDC14 (see below). The plasmid copy numbers, however, were rather firmly determined depending on the target genes cloned (Figure 2B), and the trend was related with the copy number trend in the leucine− condition (Figure S2). This indicates that there is a weak bias toward increasing the plasmid copy number due to the 2-micron plasmid, and the bias can be used to evaluate the upper limit of target genes under mild conditions. On the other hand, some genes such as BUB2, CDC15, and MIH1 interestingly showed significantly higher copy number than the vector alone (p < 0.02, two-tailed Student's t-test). Among 30 CDC genes, only CDC14 showed severe growth retardation even in the uracil− condition (CDC14 in Figure 3). To our surprise, the copy numbers of CDC14 through any leucine concentration were about 1 (CDC14 in Figure 3), and the cells even in the uracil− condition were severely elongated (unpublished data). This very low upper limit counteracting the weak bias of the 2-micron plasmid probably caused the growth retardation even in the uracil− condition.

Protein Expressed from the Target Gene Determines the Plasmid Copy Number in the gTOW Experiment

We cannot exclude the possibility that the plasmid copy number in the gTOW experiment is determined because of the effect of DNA or mRNA of the target gene, because the copy number of the plasmid DNA itself is increased in this experiment. We therefore disrupted the ORFs of target genes by inserting an adenine just after the start codon to introduce frame shift mutations. Thus, the DNA sequences of the plasmids with wild-type and the frame shift mutant were exactly the same except for one nucleotide. As shown in Figure 4A, the growth rates and the plasmid copy numbers with the frame shift mutants were increased toward the levels of the vector alone. All frame shift mutants showed more than 80 copies per haploid genome, indicating that the upper limits of genes with copy numbers of less than 80 in the leucine− condition were evaluated by the gTOW experiment due to the effect of gene products (proteins) expressed from the target genes. As mentioned above, within the genes with higher plasmid copy numbers than the vector alone in the uracil− condition, BUB2 and MIH1 showed reduced copy numbers with frame shift mutation (Figure 4B). These genes probably work positively for rapid cell growth, most likely by ignoring some checkpoints in the cell cycle.

Open in a separate window

Figure 4

Control Experiments with Frame Shift Mutants in the gTOW Experiment

(A) A scatter plot showing the correlation of maximum growth rates and plasmid copy numbers with wild-type CDC genes and their frame shift mutants determined in the gTOW experiment.

(B) Plasmid copy numbers with wild-type and frame shift mutants in the gTOW experiment in the uracil− condition. The data used in the graphs are listed in Table S2.

The Data Obtained in the gTOW Experiment Provide Unique Information

We next compared the data obtained in the gTOW experiment with other quantitative and qualitative datasets obtained in systematic analyses. First of all, the essentialities of genes [9] did not show any significant relationship with the copy number in the gTOW experiment (Spearman's r = 0.08, p > 0.05) (Figure 5A). Neither the fitness of deletion mutants of nonessential genes [9], nor haploid insufficiency [25], showed any correlation with the copy number (unpublished data). The endogenous level of proteins in wild-type cells, as determined by Ghaemmaghami et al. [26] using TAP-tagged proteins, did not show any significant correlation with the maximum tolerated plasmid copy number in our gTOW assay (Spearman's r = −0.36, p > 0.05) (Figure 5B, see also Figure S3). This indicates that the copy numbers in the gTOW experiment are determined not by the non-specific effect of overexpressed proteins to perturb general cellular functions, such as protein expression and protein degrading system, but by the specific effects of the function of each protein.

Open in a separate window

Figure 5

Comparisons of the Data Obtained in the gTOW Experiment with Other Qualitative and Quantitative Datasets

(A) The plasmid copy numbers in the gTOW experiment with essential genes and nonessential genes [9,41].

(B) A scatter plot between protein abundance obtained by Ghaemmaghami et al. [26] and the plasmid copy number in the gTOW experiment.

(C) A scatter plot of growth inhibitions between the GAL experiment and gTOW experiment with the 30 CDC genes.

(D) A scatter plot between protein abundance determined by Cross et al. [21] and the plasmid copy number of B-type cyclins in the gTOW experiment. The protein abundance is the number of copies per log-phase yeast cell of the indicated cyclin protein C-terminally tagged with protein A expressed from the endogenous promoter and chromosomal location [21].

(E) A scatter plot between selectable disadvantage determined by Cross et al. [21] and the plasmid copy number of B-type cyclins in the gTOW experiment. The selective disadvantage parameter reflects differences in doubling time between wild-type and mutant cells. The data used in the graphs are listed in Tables S1, S2, and S3.

Next, we made a comparison with the standard overexpression system using the GAL1 promoter [27]. To make the conditions the same, we constructed a series of plasmids in which each promoter on pTOW-CDCgene was replaced by the GAL1 promoter (Table 1, pGAL-CDCgene). We then measured and calculated the maximum growth rate of the cells with the plasmid. As shown in Figure 5C, in overall, both data showed significant correlation (Spearman's r = 0.82, p < 0.01). However, some genes such as CLB3, CLB5, CLB6, NET1, and PDS1 showed different degrees of inhibition between the two experiments. This is probably due to the difference of the nature of the expression system in both experiments. Thus, we found that the gTOW experiment provided unique quantitative data.

B-Type Cyclins

B-type cyclins (Clbs) were encoded in six genes (CLB1–6), which are constituted of three paralogous gene pairs (CLB1/CLB2, CLB3/CLB4, and CLB5/CLB6) [28]. One gene (CLB2, CLB3, and CLB5) or each paralogous pair is expressed stronger and has a more major function on cellular fitness. Moreover, it is reported that the fitness of deletion mutant cells of each gene and the intracellular protein level from them have a correlation [21]. As shown in Figure 5D and ​and5E,5E, CLBs, in particular, showed a close correlation between their copy numbers in the gTOW experiment with their intracellular protein abundance (Spearman's r = −0.84, p < 0.05) and selectable disadvantage (i.e., fitness of the deletion mutant) (Spearman's r = −0.77, p < 0.05). CLBs with major function (i.e., strongly expressed) had low copy numbers in the gTOW experiment, indicating that they have relatively narrow parameter ranges. Cross et al. recently measured the upper limit of Clb2 and reported it to be less than 13-fold that of the wild-type protein amount [20]. In the leucine− condition the copy number of CLB2 in the gTOW experiment was about 30 (CLB2 in Figure 3). In this condition, however, the growth of cells with the plasmid was strongly inhibited. The copy number in the condition just before the strong growth inhibition (i.e., 40 mg/ml leucine) was about 12, which is almost consistent with their finding (CLB2 in Figure 3). We also measured the Clb2 protein amount in this condition and the amount was about 12-fold that of the wild-type protein amount ( Figure 7 , see below).

Open in a separate window

Figure 7

Quantification of the Protein Level Expressed in the gTOW Experiment

(A) Typical example of the Western blots in Class I genes. Circled samples were used for the measurement of each protein bands.

(B) Typical example of the Western blots in Class II genes. Squared samples were used for the measurement of each protein bands.

(C) Scatter plot between the protein overexpression level and the gene copy number (plasmid copy number plus 1) in the gTOW experiment. The data were determined in the conditions with 100 μg/ml (CDC14), 20 μg/ml (CLB2, CLB5, and SWE1), and 30 μg/ml (CLB3) leucine conditions. The data used in the graphs are listed in Table S5.

Relationship between Upper Limit Gene Dosage and System Level Property of the Cell-Cycle System

To reveal the relationship between the copy numbers obtained in the gTOW experiment and properties of the cell-cycle system, we drew a map presenting current knowledge of molecular interactions in the cell cycle in S. cerevisiae and represented the copy number data on it (Figure S4). Interestingly, six out of seven genes with the lowest copy numbers were involved in the direct regulation of B-type CDK activity. They were the major B-type cyclin paralogous genes described above (CLB2, CLB3, and CLB5), SIC1 that encodes a stoichiometric inhibitor of B-type CDK [19], and SWE1 that encodes an inhibitory kinase of B-type CDK [23]. Moreover, it is known that they construct a subsystem with three positive feedback loops that regulate B-type CDK activity (Figure 6A), and these positive feedbacks make bi-stable B-type CDK states [23,29]. The dynamic alteration between the two B-type CDK states is thought to be the core process required for robust oscillation of the eukaryotic cell cycle [30]. Thus, it was first observed that the subsystem designed to dynamically change parameters conversely had a low limit of permissible parameter range. In contrast, G1 cyclins (CLN1, CLN2, and CLN3), CDC20, and CDH1, which alter the stable states, had higher copy numbers (>50).

Open in a separate window

Figure 6

System Level Analysis of the Copy Number in the gTOW Experiment

(A) Genes with low copy numbers in the gTOW experiment construct a subsystem that contains three positive feedback loops which regulate B-type cyclin activity [23,29]. This subsystem potentially makes two stable steady-states of B-type cyclin activity. Those states correspond to G1 phase and G2-S-M phase, respectively, which are altered by Cln-CDK and Cdc20-APC activity in the normal cell cycle [29].

(B) Comparison of the upper limit gene copy numbers obtained in the gTOW experiment and a computational cell cycle model. The upper limit gene copy numbers in vivo are considered as the plasmid copy numbers plus 1 in the gTOW experiment. Maximum fold variable change in xlog(2) toward the outside of the circle up to 256-fold is shown. The data used in the graphs are listed in Table S4.

(C) Upper limit copy numbers of stoichiometric partners in predictions of the computer model and gTOW experiment. The data used in the graph are the same as in Figure 6B.

(D) Simplified diagrams showing stoichiometric partners, where the enzymes are regulated by their inhibitors by 1:1 stoichiometric molecular interaction. The diagrams are shown as implemented in the computer model by Chen et al. [19].

Comparison of the Copy Numbers in the gTOW Experiment and a Computational Cell-Cycle Model

Chen et al. constructed a computational model in which about 30 CDC genes in S. cerevisiae regulated CDK activity [19]. The model reconstitutes 120 phenotypes out of 131 reported phenotypes and predicts quantitative behaviors of the components [19]. We compared the prediction of upper limit copy number of CDC genes in this model with the ones obtained in the gTOW experiment. For the upper limits of CDC genes in vivo, we used the plasmid copy numbers in the gTOW experiment plus one (i.e., chromosomal copy). As shown in Figure 6B, the model generally had much lower limits than in vivo results. Moreover, the model did not reproduce the system-level characteristics, that components, which directly regulate B-type cyclins (i.e., CLB2, CLB5, and MCM1), have lower limits. In fact, in the model their upper limits were rather higher than the others, although the absolute values themselves showed rather good agreement between the model prediction and in vivo data (CLB2, CLB5, and MCM1 in Figure 6B).

In the model, genes with the lowest upper limits being less than a 2-fold increase were involved in regulations by stoichiometric interactions; i.e., enzymes and their stoichiometric inhibitors (hereafter we call them “stoichiometric partners”), such as Cdc14 (protein phosphatase) versus Net1 and Esp1 (protease) versus Pds1 (Figure S7C and S7D). In the model, their regulations are easily disrupted when any of the stoichiometric partners is in relative excess more than the other, because they are regulated under the balance of 1:1 molecular interactions. Most of them did not have these extreme low limits in vivo (Figure 6C), suggesting that there are additional regulations that are not implemented into the model. However, the surprisingly low limit in CDC14 in vivo was almost perfectly consistent with the model (Figure 6C). This probably indicates that Cdc14 activity is only regulated through the stoichiometric 1:1 interaction with Net1 as it is in the model.

Plasmid constructions.

The plasmids constructed and used in this study are listed in Table 1. PCR was done with KOD-Plus high fidelity DNA polymerase (TOYOBO, Japan) according to the manufacturer's protocol. pSBI40 is a pYEX4T-1 derivative [39] (the full nucleotide sequence is available upon request). pSBI104 used in the GAL1 promoter-driven overexpression experiments was made by replacing CUP1 promoter and GST in pSBI40 by gap-repair [40] with PCR fragment containing GAL1 promoter amplified from genomic DNA using primer OSBI0246 and OSBI0247 (primer sequences are listed in Table S6B). For the gTOW, according to the Saccharomyces Genome Database (SGD) [41], DNA fragments containing target CDC genes (listed in Table 1), with upstream and downstream sequences up to their neighboring genes, were amplified from genomic DNA of BY4741 by PCR using primer sets (“up primer” and “down primer” for each gene, listed in Table S6A), then cloned into pSBI40 by the gap-repair method in BY4741. The constructed plasmids were named pTOW-target. For overexpression experiments from the GAL1 promoter, DNA fragments containing target cell-cycle genes from ATG to their downstream end were amplified from the genomic DNA by PCR using primer sets (“GAL primer” and “down primer” for each gene, listed in Table S6A), then cloned into pSBI104 by the gap-repair method in BY4741. The constructed plasmids were named pGAL-target. To make frame shift mutation, adenine was inserted just after the ATG codon of each gene except CDC14 (adenine was inserted into + 23 nt from the ATG codon), CDH1 (cytosine was inserted +128 nt from the ATG codon) and PDS1 (adenine was inserted +7 nt from the ATG codon); two PCR fragments amplified from each tug-of-war plasmid construct were amplified using: (1) “frame shift forward primer” for each gene and OSBI159 and (2) “frame shift reverse primer” for each gene and OSBI160; then they were combined and cloned into pSBI40 by gap-repair (primer sequences are listed in Tables S6A and S6C). The constructed plasmids were named pTOW-targetfs. Plasmids with more than two independent isolates from each plasmid construction were recovered from yeast, and the structure was checked by restriction digestion and partial nucleotide sequencing.

gTOW procedure.

BY4741 cells with pTOW-target containing each target gene (listed in Table 1) were grown in SC without uracil, and then they were transferred into SC without leucine and uracil. The optical density of 595 nm of the culture was monitored using a microplate reader (Model X680 Bio-Rad, Hercules, California, United States) at 30 °C without agitation every 30 min for 50 h. Mean doubling time (min) at maximum growth rate (min⁻¹) during the 150-min interval from at least three independent experiments was calculated. For the GAL experiment, BY4741 cells with pGAL-target containing each target gene (listed in Table 1) were grown in SC without uracil, before being transferred to SC without uracil with 2% galactose.

The plasmid dosage in a yeast cell was determined by two kinetic PCRs using total DNA extracted from yeast cells as a template. Yeast cells collected from 200 μl of saturated culture were suspended in lysis solution (10 mM Na-phosphate [pH 7.5], 1.2 M sorbitol, and 2.5 mg/ml Zymolyase 100T) (Seikagaku, Japan) and incubated for 10 min at 37 °C to digest the cell wall. Then the cell suspension was treated at 94 °C for 15 min, at −80° C for 5 min, then at 94 °C for 15 min. The cell suspension was chilled and centrifuged. Supernatant (containing total DNA) was used for the following two kinetic PCRs: For the kinetic PCR, we used LightCycler FastStart DNA Master^PLUS SYBER Green I (Roche, Germany) with LightCycler 2.0 instrument (Roche), according to the manufacturer's protocol. Supernatant (2 μl) was mixed with each reaction mix (18 μl) containing 0.5 μM of LEU2 (LEU2-F and LEU2-R) and LEU3 primer sets (LEU3-F and LEU3-R); primer sequences are listed in Table S6D. LEU2 and LEU3 primer sets were used to amplify and quantify LEU2 genes from the plasmid and LEU3 gene from the genome, respectively. By comparing the relative quantity of LEU2 gene and LEU3 gene, the copy number of the plasmid per haploid genome (BY4741 is a haploid strain) was estimated. The calculation is: plasmid copy number = 2^{(I.P._LEU3 - I.P_LEU2)} , where I.P._LEU3 and I.P._LEU2 are the PCR cycle numbers at inflection points of the PCR amplification curves of the LEU3 and LEU2 genes, respectively. Strictly speaking, the number determined by this procedure is not equivalent to the copy number per cell, because in G₂-M phase cells, there are two copies of genome per cell. Thus, the copy number is per haploid genome. We calculated the gene ratio of wild-type S288C strain (with LEU2 and LEU3 on the chromosome) with eight independent experiments (1.07 ± 0.16). The host strain BY4741 gave 5 × 10⁻⁴, confirming that BY4741 is a yeast strain with leu2 deletion (leu2Δ0) [24]. For each experiment in this study, we determined the copy number of more than three independent isolates from each plasmid construction, and calculated the mean value with standard deviation.

Quantification of Cdc protein overexpression.

Cells with vector pSBI40 and pTOW-target were cultivated in 2 ml SC without uracil and 2 ml SC without leucine (otherwise stated) for two overnights, then 4 ml of the fresh medium were added to the culture which was then cultivated for 4 more h to refresh the cells. DNA was extracted from the cells and the plasmid copy number was determined. Protein extraction was performed as described [42]. Briefly, cells from each 2-ml culture were collected and suspended in 400 μl of 0.2 N NaOH; after 5 min of incubation at room temperature, cells were collected and re-suspended in 100 μl of SDS-sample buffer [42], then heated at 100 °C for 5 min. Cell debris was removed by centrifugation, then the supernatant with indicated dilution was separated in 10% SDS-PAGE for the standard Western blot procedure.

In immunoblotting, each target Cdc protein was detected using its specific antibody (listed in Table 2, obtained from Santa Cruz, California, United States) with 1/200 dilution, and anti-Goat IgA peroxydase conjugate (A5420, Sigma, St. Louis, Missouri, United States) with 1/5000 dilution. Hexokinase was detected using anti-hexokinase antibody (100-4159, Sigma) with 1/1000 dilution, and anti-rabbit IgG, IgM, IgA HRP conjugate (SAB1003, Open Biosystems, United States) with 1/5000 dilution. Detection of immune complex was done with SuperSignal West Dura Extended Duration Substrate (34075, Pierce Biotechnology, Rockford, Illinois, United States). Intensities of corresponding protein bands were measured using an LAS-3000 mini image analyzer (Fuji Film, Japan), and the data within the linear detection range among the serial dilution of the samples were used.

If the endogenous Cdc protein was detected (we classified the antibody and the detected protein as Class I), the Cdc protein overexpression was quantified as the fold increase of the target protein over the endogenous protein amount (i.e., vector alone), which was normalized using hexokinase as a standard of total protein amount. The calculations are: Cdc overexpression in uracil− = (Cdc_{target_ura} / Cdc_{vec_ura}) / (Hxk_{target_ura} / Hxk_{vec_ura}), Cdc overexpression in leucine− = (Cdc_{target_leu} / Cdc_{vec_ura}) / (Hxk_{target_leu} / Hxk_{vec_ura}), where Cdc and Hxk mean the intensity of the target protein and hexokinase bands in the Western blot of the samples from indicated plasmid and culture conditions described in subscript (i.e., target; pTOW-target, vec: vector, ura: uracil− condition, leu: leucine− condition). If the endogenous Cdc protein was not detected, but the protein overexpressed from the pTOW-target was detected (we classified the antibody and the detected protein as Class II), we estimated the least-fold increase of the target protein using the number of the highest dilution in which the Cdc protein was detectable which was normalized using hexokinase. The calculations are: Least Cdc protein overexpression in uracil− = MD / (Hxk_{target_ura} / Hxk_{vec_ura}), Least Cdc protein overexpression in leu2− = MD × (Cdc_{target_leu} / Cdc_{target_ura}) / (Hxk_{target_leu} / Hxk_{vec_ura}), where MD is the maximum dilution number of the sample from the target_ura condition, in which the target Cdc protein band in Western blot could be detectable. At least two independent experiments were done for each target protein, and representative data were shown.

The authors would like to thank members of the Systems Biology Institute (SBI) for their helpful discussions and Takuya Maeda for assisting with the computer simulations.