1. Genetics and Genomics
Download icon

Genetic profiling of protein burden and nuclear export overload

  1. Reiko Kintaka
  2. Koji Makanae
  3. Shotaro Namba
  4. Hisaaki Kato
  5. Keiji Kito
  6. Shinsuke Ohnuki
  7. Yoshikazu Ohya
  8. Brenda J Andrews
  9. Charles Boone
  10. Hisao Moriya  Is a corresponding author
  1. Donnelly Center for Cellular and Biomolecular Research, Department of Medical Genetics, University of Toronto, Canada
  2. Research Core for Interdisciplinary Sciences, Okayama University, Japan
  3. Matching Program Course, Okayama University, Japan
  4. Graduate School of Environmental and Life Science, Okayama University, Japan
  5. Department of Life Sciences, School of Agriculture, Meiji University, Japan
  6. Graduate School of Frontier Sciences, University of Tokyo, Japan
  7. RIKEN Center for Sustainable Resource Science, Japan
Research Article
  • Cited 0
  • Views 1,318
  • Annotations
Cite this article as: eLife 2020;9:e54080 doi: 10.7554/eLife.54080

Abstract

Overproduction (op) of proteins triggers cellular defects. One of the consequences of overproduction is the protein burden/cost, which is produced by an overloading of the protein synthesis process. However, the physiology of cells under a protein burden is not well characterized. We performed genetic profiling of protein burden by systematic analysis of genetic interactions between GFP-op, surveying both deletion and temperature-sensitive mutants in budding yeast. We also performed genetic profiling in cells with overproduction of triple-GFP (tGFP), and the nuclear export signal-containing tGFP (NES-tGFP). The mutants specifically interacted with GFP-op were suggestive of unexpected connections between actin-related processes like polarization and the protein burden, which was supported by morphological analysis. The tGFP-op interactions suggested that this protein probe overloads the proteasome, whereas those that interacted with NES-tGFP involved genes encoding components of the nuclear export process, providing a resource for further analysis of the protein burden and nuclear export overload.

Introduction

Extreme overproduction of a gratuitous protein that has no cellular function causes growth defects, which, at least in part, appears to be caused by overloading the cellular resources for protein synthesis (Dong et al., 1995; Snoep et al., 1995; Stoebel et al., 2008; Makanae et al., 2013; Shah et al., 2013; Kafri et al., 2016; Moriya, 2015; Eguchi et al., 2018; Scott et al., 2010). This phenomenon is called the protein burden/cost and has been extensively studied in the budding yeast Saccharomyces cerevisiae, a model eukaryotic cell. Limiting functions defining the protein burden are thought to be the translational process upon nitrogen limitation, and the transcriptional process upon phosphate limitation (Kafri et al., 2016). The protein burden itself initially appears to be a relatively simple phenomenon, but little is known about the physiological conditions and cellular responses triggered by the protein burden.

To trigger the protein burden, a protein must be produced at a level sufficient to overload protein production resources (Moriya, 2015; Eguchi et al., 2018). This can happen only if the protein is otherwise harmless. Fluorescent proteins, such as EGFP, Venus, and mCherry, do not have any physiological activity in yeast cells and thus are considered gratuitous proteins. Therefore, these fluorescent proteins are believed to be produced at the highest possible levels in yeast cells, and their overproduction triggers a protein burden (Makanae et al., 2013; Kafri et al., 2016; Eguchi et al., 2018; Farkas et al., 2018). Modifications to EGFP, such as adding a degradation signal, misfolding mutations, or adding localization signals, reduces its expression limit, probably because these modifications overload limited resources for the degradation, folding, and localization processes, respectively (Geiler-Samerotte et al., 2011; Makanae et al., 2013; Kintaka et al., 2016; Eguchi et al., 2018).

A recent study isolated a group of deletion strains in which growth defects upon overproduction of yEVenus are exacerbated (Farkas et al., 2018). Through the analysis of these strains, and conditions exacerbating the protein burden, the authors concluded that Hsp70-associated chaperones contribute to the protein burden by minimizing the damaging impact of the overproduction of a gratuitous protein. Chaperone genes, however, constitute only a relatively small fraction of the deletion strains isolated in the study, and thus the protein burden may impact numerous other processes.

To understand the physiological conditions caused by protein burden, we conducted a systematic survey of mutants that exacerbate or alleviate the growth inhibition caused by GFP overproduction. We surveyed genetic interactions between mutant strains and high levels of GFP overproduction (GFP-op) to genetically profile cells exhibiting this phenomenon. Here, if a mutation exacerbates growth inhibition by GFP-op, or if GFP-op exacerbates growth inhibition by the mutation, the mutation has a negative genetic interaction with GFP-op. Also, if a mutation alleviates growth inhibition caused by GFP-op, the mutation has a positive genetic interaction with GFP-op. If GFP-op relaxes the growth inhibition caused by the mutation, it is also detected as a positive genetic interaction.

To isolate mutant sets showing positive and negative genetic interactions with the protein burden, we used a condition causing significant growth defects due to high GFP-op from the TDH3 promoter (TDH3pro) on a multi-copy plasmid. In addition to a deletion mutant collection of non-essential genes, we surveyed temperature-sensitive (TS) mutant collections of essential genes. We performed a strict statistical evaluation to isolate mutants showing robust genetic interactions with high confidence.

We also attempted to distinguish between the protein burden and other process overloads by surveying genetic interactions between those mutant strains and a triple-GFP (tGFP) with a nuclear export signal (NES-tGFP). NES-tGFP triggers growth defects at a lower expression level than unmodified tGFP (Kintaka et al., 2016). If the protein burden can only be triggered by a harmless protein like GFP, mutants harboring genetic interactions with tGFP-op should be different from those with NES-tGFP-op, and the comparison of those mutants will identify consequences specific to the protein burden. Moreover, mutants harboring negative genetic interactions should contain limiting factors of the nuclear export and essential factors affected by the overloading of nuclear export.

Results

Isolation of mutants that have genetic interactions with GFP-op

To isolate mutants genetically interacting with GFP-op, we performed a synthetic genetic array (SGA) analysis (Baryshnikova et al., 2010Figure 1A). As a query strain, we overproduced GFP (yEGFP) (Cormack et al., 1997) under the control of TDH3pro on the multi-copy plasmid pTOW40836 (Figure 1B). This plasmid contains two selection markers (URA3 and leu2-89), and the copy number can be controlled by the culture conditions. The copy numbers of this plasmid under –Ura and –Leu/Ura conditions are around 10 and 30 copies per cell, respectively (Eguchi et al., 2018). While a strain harboring this plasmid shows growth defects even under –Ura conditions (Figure 1C), the strain shows more growth defects under –Leu/Ura conditions (Figure 1D), presumably because the copy number increase leads to an increase in GFP production, and probably causes a stronger protein burden-associated growth defect (Eguchi et al., 2018). The background principles that determine plasmid copy number and growth rate (genetic tug-of-war) are explained in detail in Figure 1—figure supplement 1.

Figure 1 with 2 supplements see all
Experimental scheme of genetic interaction (GI) analysis.

(A) Each mutant from a deletion mutant array (DMA) and a temperature-sensitive mutant array (TSA) was combined with GFP overproduction (GFP-op) using the synthetic genetic array (SGA) method (Baryshnikova et al., 2010). The colony size of each derivative strain grown on synthetic complete (SC)–Ura and SC–Leu/Ura plates was measured to calculate a genetic interaction (GI) score (ε). Four colonies were analyzed for each strain, and the entire experiment was duplicated. (B) The structure of the plasmid used to overexpress GFP. The plasmid copy number, and thus the expression level of GFP, can be changed by changing the growth conditions. (C and D) Effect of GFP production on growth under each condition. The size of colonies of each strain grown on agar medium was measured (n > 12). The Y7092 strain was used as the host. The maximum growth rate was measured in liquid culture (n > 4). The average, standard deviation (error bar), and p-value of Student’s t-test are shown.

We examined an array of 4323 deletion mutants in nonessential genes (DMA) and an array of 1016 conditional temperature-sensitive mutants (TSA) (Costanzo et al., 2016). Details on the calculation of genetic interaction scores from colony size are shown in Figure 1—figure supplement 2. We assessed the growth (fitness) of wild-type and mutant strains by measuring the colony size on agar plates under vector control and GFP-op conditions, respectively. This colony size was then normalized by the overall colony size and the relative fitness of each strain was obtained as the normalized colony size. For each mutant strain, we calculated genetic interaction (GI) scores (ε) from the analysis of four colonies under both –Ura and –Leu/Ura conditions, in duplicate (Figure 2—source data 1). After thresholding by the variation in colony size (p<0.05), we compared GI scores between duplicates (Figure 2A, Figure 2—figure supplement 1). The reproducibility of the DMA experiments was lower in –Ura conditions (r = 0.17), whereas it was higher in –Leu/Ura conditions (r = 0.36). The reproducibility of the TSA experiments was higher in both –Ura and –Leu/Ura conditions (r = 0.42 and 0.53). Thus, the conditions which cause severe growth defects produce the most reproducible GI scores.

Figure 2 with 5 supplements see all
Characteristics of GI scores.

(A) Pearson correlation coefficient (r) of GI scores from experimental duplicates. DMA and TSA: comparison of all GI scores of duplicates obtained by the GI analysis using DMA and TSA. DMA-0.08 and TSA-0.08: comparison of GI scores of duplicates with value > |0.08| obtained by the GI analysis using DNA and TSA. Figure 2—figure supplement 1 shows an independent comparison. (B and C) Comparison of average GI scores of DMA (B) and TSA (C) mutants both with GI scores in the duplicates > |0.08| under –Ura and –Leu/Ura conditions. (D) GI score (ε) of mutants isolated ordered by score ranking. Mutants with low (<0.2) and high (>0.2) scores are shown in light blue and orange, with enriched GOs in those mutants. The score in –LU is shown. The full list of enriched genes is in Supplementary file 1.

To more confidently identify mutants showing strong GIs, we set a threshold in each replicate (ε > |0.08|). Previous studies have reported that the use of these thresholds results in more reproducible genetic interactions (Baryshnikova et al., 2010). Using this threshold increased reproducibility, especially in the DMA experiments (r = 0.35 in –Ura, r = 0.62 in –Leu/Ura, Figure 2A). We first selected mutants with ε > |0.08| in each replicate and then calculated their average GI scores between the duplicates as summarized in Figure 2—figure supplement 2. Because GI scores between –Ura (low-level GFP-op) and –Leu/Ura (high-level GFP-op) conditions were highly correlated (r = 0.70 and 0.58, Figure 2B and C), this procedure identified high-confidence mutants with GIs with GFP-op. We note that there is a higher correlation between conditions at –Ura and –Leu/Ura than between replicates in the DMA experiment (Figure 2A and Figure 2B). The cause of this is unclear, but it may indicate that averaging between replicates yields values closer to the true GI score.

Farkas et al. surveyed GIs between deletion mutants and the overproduction of yEVenus (Farkas et al., 2018). The GI scores obtained by our analysis did not show correlation with those from the Farkas study (r = –0.01 and –0.07, Figure 2—figure supplement 3A,B). This may be because of the weak reproducibility observed in lower overproduction conditions (Figure 2—figure supplement 1A). Moreover this overlap analysis only involved nonessential genes and the Farkas study used a relatively weaker HSC82 promoter (HSC82pro), in medium comparable to our –Ura condition, in which the GFP-op from HSC82pro on pTOW40836 caused a very minor growth defect in –Ura conditions (Figure 2—figure supplement 3E,F). Indeed, our conditions produced more variance in the GI scores and thus identified more mutants showing stronger GIs (Figure 2—figure supplement 3A,B), and we found that negative GIs of 6 out of 7 deletion mutants from our screening were confirmed by independent growth measurements in the liquid medium, while all six mutants isolated by the previous study (Farkas et al., 2018) were not (Figure 2—figure supplement 4). Farkas et al. reported that growth inhibition (cost) due to overproduction of yEVenus was stronger as the concentration of amino acids in the medium decreased. As shown in Figure 2—figure supplement 3G, overproduction of yEGFP also resulted in increased growth inhibition (cost) due to amino acid dilution, although the degree of cost differed between the two fluorescent proteins. We cannot dismiss the possibility that the difference between the analysis of Farkas et al. and ours is due to the difference in properties between our GFP (yEGFP) and yEVenus.

During the screening, we noticed that a group of temperature-sensitive mutants showed greater growth defects under –Leu/Ura conditions than under –Ura conditions in the vector control experiments (Figure 2—figure supplement 5). The gene ontology (GO) term ‘DNA replication preinitiation complex [GO:0031261]’ was significantly over-represented in the mutated genes (seven genes, p=1.47E–05). Figure 2—figure supplement 5A shows the normalized colony size differences of the 18 mutants analyzed in the TSA corresponding to the genes categorized in GO:0031261. 6 out of 18 mutants showed more than 2U decrease in their colony sizes, whereas the average of all temperature-sensitive mutants showed 0.002U (Rep1) and 0.003U (Rep2). Colonies of representative mutants (cdc47-ts, orc1-ph, and orc6-ph) are shown in Figure 2—figure supplement 5B. The vector copy number is more than 100 copies per cell under –Leu/Ura conditions (Makanae et al., 2013; Eguchi et al., 2018). This high copy number probably produces limitations of the replication initiation complex by sequestering the complex to the replication origins of the plasmids (the explanation of the fitness reduction is shown in Figure 2—figure supplement 5C). Some negative factors on the plasmid, like TDH3pro-GFP, restrict the plasmid copy number due to a genetic tug-of-war effect (Figure 1—figure supplement 1CMoriya et al., 2006), and the plasmid thus may not trigger the limitation of the replication initiation complex. This situation may lead to a bias toward the isolation of mutants in the replication initiation complex with positive GIs with plasmids containing toxic elements, especially under –Leu/Ura conditions.

Mutations aggravating or mitigating GFP-op triggered growth defects

To understand which processes are affected by GFP-op, we performed enrichment analysis targeted toward isolating mutants with stronger GIs (ε > |0.2|) under –Leu/Ura conditions, as the results obtained under these conditions were more reproducible (Figure 2A, Figure 2—source data 1). Therefore, we believed that stronger and more confidnet genetic interactions could be obtained with this threshold. We designated the negatively interacting genes and mutants ‘GFP-op_negative’ and the positively interacting genes and mutants 'GFP-op_positive'. The GFP-op_negaitive 71 genes (79 mutants) were significantly enriched in GO categories related to cytoskeletal organization and polarization (Figure 2D, Supplementary file 1). Figure 3A shows the GI scores under –Leu/Ura conditions of all 45 alleles of the GFP-op_negative genes categorized in GO as 'cellular bud [GO:0005933]'. Most of the mutants showed negative GIs, and 16 out of 45 showed average scores of less than –0.2.

Independent GI scores ( ε) of genes enriched in GO categories in GFP_negative and GFP_positive genes.

(A) GI scores of mutants isolated as GFP_negative genes annotated with the GO term 'cellular bud [GO:0005933]'. (B) GI scores of mutants annotated with the GO terms ‘TRAMP complex [GO:0031499]’ and 'nuclear exosome [GO:0000176]'. (C) GI scores of mutants annotated with the GO term 'Mediator-RNA polymerase II preinitiation complex [GO:0090575]'. GI scores from experimental replicates under –Leu/Ura conditions are shown. Temperature-sensitive mutant of essential genes and deletion mutant of non-essential gene are shown in different colors. GFP levels for each strain are also shown.

One hundred GFP-op_positive genes (100 mutants) were enriched in genes involved in RNA 3′-end processing and the transcription factor complex (Figure 2D, Supplementary file 1). Among the factors in the RNA 3′-end processing, the subunits in the ‘TRAMP complex [GO:0031499]’ and ‘nuclear exosome [GO:0000176]’ were isolated as GFP-op_positive genes. Figure 3B shows the GI scores under –Leu/Ura conditions of the mutants of the TRAMP complex and the nuclear exosome subunits. Of 13 mutants, seven showed positive GIs with average scores greater than 0.2. Among the transcription factor complex, subunits of the ‘mediator-RNA polymerase II preinitiation complex [GO:0090575]’ were specifically isolated. Figure 3C shows the GI scores under –Leu/Ura conditions of the mutants of the mediator-RNA polymerase II preinitiation complex subunits. In total, 20 out of 38 mutants showed positive GIs with average scores greater than 0.2.

Investigation of GFP expression levels of mutants

We next investigated GFP expression levels of each mutant overexpressing GFP. To obtain the GFP expression level of each mutant, we measured normalized GFP fluorescence (GFPunit) from the fluorescence intensity of each colony (Figure 4A). As summarized in Figure 4B, the GFPunit can be used to interpret the mechanisms underlying GFP-op_negative and GFP-op_positive mutations as follows: (1) if GFPunit is lower in a GFP_negative mutant, the mutant is considered to be more sensitive to GFP overproduction; (2) if GFPunit is higher in a GFP_negative mutant, the mutant triggers elevated GFP production, which potentially enhances protein burden; (3) if GFPunit is lower in a GFP_positive mutant, the mutant triggers reduced GFP production, which potentially mitigates protein burden; and (4) if GFPunit is higher in a GFP_positive mutant, the mutant is considered to be resistant to GFP overproduction. The detailed background principle is explained in Figure 4—figure supplement 1.

Figure 4 with 3 supplements see all
Experimental scheme of GFP expression measurements of mutants.

(A) Each mutant from a deletion mutant array (DMA) and a temperature-sensitive mutant array (TSA) was combined with GFP overproduction (GFP-op) with background E2-Crimson expression, using a synthetic genetic array (SGA) method. The median GFP fluorescence (F488) and E2-Crimson fluorescence (F532) of each colony were measured, and the GFP expression level (GFPunit) of each mutant was calculated by dividing F488 by F532 to normalize colony size. (B) Further classification and interpretation of mutant strains exhibiting positive/negative genetic interactions using GFP levels. A more detailed explanation is shown in Figure 4—figure supplement 1. (C) GFPunits of GFP-op_negative mutants. Mutants with lower and higher GFPunits than the average are designated as GFP_L and GFP_H mutants, respectively. Representative GO terms enriched in GFP_L mutants in GFP-op_negative mutants are shown. (D) GFPunits of GFP-op_positive mutants. Representative GO terms enriched in GFP_L mutants and GFP_H mutants in GFP-op_positive mutants are shown. The full list of enriched genes is in Supplementary file 2A.

Of the mutants, 1447 (29%) showed lower GFPunits and 3572 (71%) mutants show higher GFPunits than the average of all mutants (Figure 4C,D, Figure 4—source data 1). We designated these mutants GFP_H and GFP_L, respectively (Figure 4C,D). Mutants with enhanced growth defects upon GFP overproduction (GFP-op_negative mutants) were more likely to produce less GFP (GFP_L) (Figure 4C, p=4.7E-11, Student’s t-test), indicating that the limit of GFP overproduction in these cells was lower than in other cells. Eleven out of 13 GFP-op_negative mutants categorized as ‘cellular bud [GO:0005933]’ were also GFP_L (Figure 4—figure supplement 2A, Supplementary file 2A) (the GFP levels for each strain in this category are shown in Figure 3A). These mutants seemed to be sensitive to the protein burden.

In contrast, only a slightly higher number of mutants in which the growth inhibition caused by GFP overproduction was alleviated (GFP-op_positive) had lowered GFP expression (GFP-L) (Figure 4D, p=0.013, Student’s t-test). Trends in the distributions of mutants in 'TRAMP complex [GO:0031499]', 'nuclear exosome [GO:0000176]', and ‘mediator-RNA polymerase II preinitiation complex [PMID27610567]’ were not obvious (Figure 4—figure supplement 2B, Supplementary file 2A; the GFP levels for each strain in these categories are shown in Figure 3). However, GFP-op_positive and GFP_L mutants were significantly enriched in 'RNA polymerase II transcriptional factor complex [GO:0090575]', suggesting that these mutants may simply cause the reduction of GFP production, but not decrease the sensitivity to the protein burden.

The GFP-op_positive and GFP_H strains include strains resistant to GFP overproduction (i.e. protein burden). We recently reported one such strain, a deletion of the dubious gene YJL175W, which produces a partial deletion of SWI3 (Saeki et al., 2020). We next searched for mutations where the mutation creates a growth advantage and where this advantage is further enhanced by GFP overproduction. We identified 14 mutants among GFP-op_positive and GFP-H mutants whose fitness in the vector control was higher than that of other strains and whose fitness may be further enhanced by GFP overproduction (Figure 4—figure supplement 3). The 14 mutants were significantly enriched in genes of the DASH complex [GO:0042729] (Supplementary file 3), including ask1-2, dad2-9, and spc34-5 out of the seven DASH complex mutants analyzed (Figure 4D, Figure 4—figure supplement 3C,D). The DASH complex binds to microtubules and is involved in the distribution of chromosomes (Jenni and Harrison, 2018). At present, the molecular mechanism by which GFP overproduction is permissible in these mutants cannot be readily deduced and more detailed analysis is required.

Overproduction of tGFP and NES-tGFP results in GIs with distinct sets of genes

We next analyzed mutants genetically interacting with a GFP containing a nuclear export signal (NES). Instead of GFP, we used tGFP made from three linked GFPs (Figure 5A) for the following reasons. In a previous study, we found that the addition of NES to monomeric GFP with a molecular weight smaller than the exclusion limit of the nuclear pore does not localize outside the nucleus, and when the molecular weight is increased by linking three GFPs together, the extranuclear localization of tGFP is clearly established (Kintaka et al., 2016). In addition, probably because NES-GFP undergoes repeated free transport into the nucleus and transport out of the nucleus by means of transport machinery, overproduction of NES-GFP shows a very strong growth inhibition. And we have found that this growth inhibition is mitigated to some extent by using tGFP (Kintaka et al., 2016). If the growth inhibition is too strong, we cannot generate overexpressing mutants to detect genetic interactions. We used NES from PKI, and used the PYK1 promoter, because the TDH3 promoter is too strong and causes severe growth inhibition (data not shown). We also used PYK1pro-tGFP as a control for NES-tGFP (Figure 5A). Using the same procedures as in the analysis of GFP described above except upper and lower threshold of ε 0.16 and –0.12 as these thresholds have been used to obtain confident genetic interactions in previous studies (Costanzo et al., 2010; Costanzo et al., 2016), we isolated total 714 mutants (695 genes) harboring GIs with either GFP-op, tGFP-op or NES-tGFP-op under –Leu/Ura conditions (the raw data sets are in Figure 5—figure supplement 1—source datas 1 and 2, the isolated mutants are in Figure 2—source data 2). To extract genes that had specific GIs with each condition, we performed clustering analysis using them, which were isolated in at least one of GFP-op, tGFP-op, and NES-tGFP experiments (Figure 5C, Figure 5—source data 1).

Figure 5 with 2 supplements see all
GFP-op harbor GIs with distinct sets of genes from those with tGFP-op and NES-tGFP-op.

(A) Structures and promoters used to overexpress GFP, tGFP, and NES-tGFP. Nucleotide sequences of the three GFPs in tGFP (and NES-tGFP) are different, other to avoid recombination. (B) Colony size of query strains with the vector control and overproduction plasmids. The size of colonies of each strain grown on –Leu/Ura agar medium was measured (n > 6). The Y7092 strain was used as the host. The average, standard deviation (error bar), and p-value of Student’s t-test are shown. (C and D) Clustering analysis of the mutants having GIs with GFP-op, tGFP-op, and NES-tGFP-op (C), and its characterization (D). Total 714 mutants (695 genes) with upper and lower threshold of ε 0.16 and –0.12 harboring GIs with either GFP-op, tGFP-op or NES-tGFP-op under –Leu/Ura conditions were used (Figure 5—source data 1). (E) GFPunits of NES-tGFP-op_negative mutants. Mutants with lower and higher GFPunits than the average are designated as NES-tGFP_L and NES-tGFP_H mutants, respectively. Representative GO terms enriched in GFP_L mutants in GFP-op_negative mutants are shown. Nuclear transport mutants obtained in Cluster 3 in D are also shown on the graph. The full list of enriched genes is in Supplementary file 2B.

Figure 5D shows the representative GO term or publication for each cluster (the whole data is shown in Supplementary file 4). Mutants negatively interacting only with NES-tGFP-op (Cluster 3) contained mutants of genes playing a central role in the nuclear protein export (Crm1, Gsp1, Rna1, and Yrb1). GI scores of these mutants were significantly lower in the NES-tGFP-op experiment than in the other two experiments (Figure 5—figure supplement 1A), suggesting that NES-tGFP-op specifically causes growth defects through overloading these limited factors.

To our surprise, only 12% (81/688) of mutants showed shared GIs between GFP and tGFP. Mutants negatively interacting with tGFP-op and NES-tGFP-op but not GFP-op (Cluster 4) were strongly enriched in annotations of ‘cytosolic proteasome complex [GO:0031597]’ (Figure 5D). GI scores of mutants in ‘proteasome complex [GO:0000502]’ were significantly lower in the tGFP-op and NES-tGFP-op experiments than in the GFP-op experiment (Figure 5—figure supplement 1B). These results suggest that GFP and tGFP have different characteristics, and tGFP-op triggers proteasome stress.

Mutants interacting only with GFP-op (Cluster 6 and Cluster 11) were enriched in genes annotated to ‘cellular bud neck [GO: 0005935]’ and 'transcription by RNA polymerase II [GO:0006366]', and their GI scores were significantly lower and higher in tGFP-op experiments than in the other two experiments (Figure 5—figure supplement 1C,D). This observation suggests that these two processes are specifically interacting with the protein burden and can be only triggered by proteins with very high expression.

We also measured tGFP and NES-tGFP expression levels in tGFP-op and NES-tGFP-op experiments using the same method as shown in Figure 4, and the results are shown in Figure 5—figure supplement 2. Among the mutants that had a negative genetic interaction with NES-tGFP-op, the mutants with lower GFP expression (NES-tGFP_L) were enriched for genes involved in nuclear protein transport (Figure 5E). Furthermore, all mutants in Cluster 3 in Figure 5D (crm1-1, gsp1-162, rna1-1, rna1-s116f, and yrb1-52) were included in this group (Figure 5E, Supplementary file 2B). As explained in Figure 4—figure supplement 1B, this result implies that these mutants are sensitive to high expression of NES-tGFP. The identification of nuclear protein transport mutants as specifically sensitive to overproduction of NES-tGFP can be considered as a proof of concept for this study, and the results strongly support that the analysis of GFP-op and tGFP-op also allowed us to obtain mutants specifically associated with the physiological state of the cells triggered by the overproduction of them.

Overproduction of triple GFP causes the formation of ubiquitinated intracellular condensates, which in turn may overload the proteasome

As noted above, we used tGFP as a control for NES-tGFP. We initially expected that tGFP was simply a protein consisting of three molecules of GFP linked together and having the same properties as monomeric GFP, and that their overproduction would show a similar genetic interaction profile. Because the amount of GFP expressed from one copy of tGFP is three times greater than that of GFP, we also predicted that the expression limit of tGFP would simply be reduced to one-third of the number of molecules as GFP. However, as mentioned above, the mutants that show a genetic interaction between the two were very different, suggesting that overproduction of tGFP causes a completely different effect than overproduction of GFP. In particular, as described above, only overproduction of tGFP (and NES-tGFP) showed a negative genetic interaction with the proteasome mutants (Figure 5D, Cluster 4).

Therefore, we next analyzed the properties of tGFP in more detail. Figure 6A shows the expression levels of tGFP in mutant strains that show negative genetic interactions with tGFP. Proteasome mutants were identified in both low (tGFP_L) and high (tGFP_L)-tGFP expression mutants. This suggests that overexpressed tGFP is actively degraded by the proteasome in the wild-type strains, whereas overaccumulation of undegraded tGFP occurs in proteasome mutations or that undegraded tGFP is cytotoxic in mutations of the proteasome.

Figure 6 with 1 supplement see all
Overexpression of triple GFP causes the formation of intracellular condensates, which in turn may overload the proteasome.

(A) GFPunits of tGFP-op_negative mutants. Mutants with lower and higher GFPunits than the average are designated as tGFP_L and tGFP_H mutants, respectively. Representative GO terms enriched in GFP_L mutants in GFP-op_negative mutants are shown. Proteasome mutants (excluding allele names) are also shown on the graph. The full list of enriched genes is in Supplementary file 2C. (B) Quantification of expression limits of GFP, tGFP, and NES-tGFP. Western blot analysis of total protein from GFP-op (1/10 diluted), tGFP-op, and NES-tGFP cells cultured in SC–Ura medium. Relative GFP levels (protein units) were calculated by measuring the intensities of bands corresponding to the molecular weight of each protein (arrowheads). Note that molar concentration GFP should be divided by three in the case of tGFP and NES-tGFP because they have three times more epitopes for the antibody than GFP. (C) Microscope images of cells overexpressing GFP, tGFP, and NES-tGFP. The nucleus was observed using Hoechst 33342 staining. Representative cells with intracellular condensates are indicated by green arrowheads (condensates with GFP fluorescence) and yellow arrowheads (nucleus). (D) Analysis of GFP and tGFP aggregation. Extracts of yeast grown in SC–Leu/Ura medium were centrifuged, fractionated into supernatant (Sup) and precipitate (Ppt), and separated by SDS-PAGE. Electrophoretic images of all proteins and western blots with anti-GFP antibodies are shown. The molecular weights corresponding to GFP and tGFP are indicated by arrows. The amounts of GFP and tGFP detected in the supernatant and precipitation are shown with GFP in the supernatant as 100. The percentages of precipitation to the total are also shown. (E) Detection of proteins in precipitation fractions. Proteins in the precipitated fractions of the vector control and tGFP-op cells were detected by LC-MS/MS and their amounts were compared by peptide-spectrum match (PSM). Three proteins that were particularly abundant in the precipitation fractions of tGFP-op cells were indicated on the plot. The components of the proteasome (Ppn, Prt, and Pre proteins) are shown by red circles. The average values of three LC-MS/MS measurements are shown. The raw data are shown in Figure 6—source data 1.

If tGFP is actively degraded, then the amount of tGFP in the cell would be much lower than that of GFP. Indeed, when we investigated the abundance of GFP and tGFP by Western blotting, the amount of tGFP (and NES-tGFP) was about 4% of GFP (Figure 6B). As mentioned above, if GFP and tGFP have the same properties and generate protein burden in the same way, they should express about the same amount of protein units and about one-third of the number of molecules as GFP. Note that under these conditions (–Ura), where the protein levels were measured, cells harboring GFP, tGFP, and NES-tGFP plasmids showed a significant delay in the growth rate compared to the vector control (Figure 6—figure supplement 1A). From this, it can be said that the expression levels of these proteins have a negative effect on cell proliferation. Thus, tGFP (and NES-tGFP) was found to have different properties than GFP, as overproduction of only a few percent of GFP may trigger growth inhibition.

Next, we analyzed the behavior of overproduced tGFP in yeast cells. First, the subcellular localization of GFP, tGFP, and NES-tGFP was observed by fluorescence microscopy. We found that giant condensate were present only in the cells overproducing tGFP and NES-tGFP (Figure 6C). To further investigate the nature of this structure, cell fractionation was performed; the cell lysate was separated into soluble supernatant and insoluble precipitate. Fluorescence microscopy showed that the aggregates seen in tGFP-overexpressing cells were enriched in the precipitation (Figure 6—figure supplement 1B). Proteins in each fraction was then separated by SDS-PAGE and detected for GFP by Western blotting (Figure 6D). About 1% of GFP was found in the precipitation fraction, whereas 14% of tGFP was found in the precipitation fraction. Western blotting with anti-ubiquitin antibodies detected ubiquitinated high-molecular-weight proteins in the precipitates of tGFP-op (Figure 6—figure supplement 1C). We next identified the proteins in the precipitation by liquid chromatography-tandem mass spectrometry (LC-MS/MS). More than 1000 proteins were detected in the precipitation of the vector control and tGFP-op (Figure 6E). Among them, Hsp70 (Ssa1/Ssa2) and the glycolytic enzymes Fba1 and Eno2 were particularly abundant in the precipitates of tGFP-op. Components of the proteasome were also identified, albeit in trace amounts, and tended to be more abundant in the precipitation of tGFP-op than in the vector control. Ubiquitin was not detecte in this experiment.

GFP-op affects actin distribution

The above results indicate that GFP-op, that is the protein burden, could affect actin functions. We thus performed a morphological analysis of cells under GFP-op with a high-throughput image-processing system (CalMorph) (Ohtani et al., 2004). We used non-fluorescent GFP mutant (GFPy66g) for this analysis because strong GFP fluorescence affects the observation of the cell shape with FITC-ConA. We also analyzed the cells overexpressing Gpm1 and a catalytically negative Gpm1 mutant (Gpm1-m) whose overproduction is considered to cause the protein burden (Eguchi et al., 2018). Cells were cultured under SC–Ura conditions. Among obtained 501 morphological parameters, only four parameters showed significant differences over the vector control, and three of them (A120_A1B, ACV7-1_A, and A122_A1B) were actin-related parameters (Figure 7A–D). Figure 7E shows the interpretation of the morphology of GFP-op cells. The cells contained increased actin patch regions, supporting the idea that the protein burden interacts with actin function.

Figure 7 with 1 supplement see all
Morphological analysis of the cells overexpressing gratuitous proteins, and models explaining the consequence of overexpression.

(A–D) Morphological parameters significantly different all in the cells overexpressing GFPy66g, Gpm1, and Gpm1-m cells over the cells with the vector control. *: FDR = 0.01 by Wald test. To overexpress GFPy66g, Gpm1, and Gpm1-m, pTOW40836-TDH3pro-GFPy66g, pTOW40836-TDH3pro-Gpm1, and pTOW40836-TDH3pro-Gpm1-m were used. (E) Interpretation of the morphology of GFP-op cells according to the morphological parameters significantly different from the vector control. (F) Dissection of the consequence of protein overproduction by the expression limits. Only otherwise harmless protein could cause the protein burden, which is associated with the perturbation of actin function. (G) A ‘barrel model’ to explain the relationship between the capacity of intracellular processes and the limits of protein synthesis. An explanation of this model is described in Discussion.

Discussion

In this study, we genetically profiled the consequences of protein overproduction using GFP as a model gratuitous protein and NES-tGFP as a transported model protein. We confirmed our prediction that the overproduction of NES-containing protein (NES-tGFP) overloads the amount of limiting nuclear-export factors (Kintaka et al., 2016). Overproduction of NES-tGFP had strong negative GIs with mutants in the major nuclear export factors (Crm1, Gsp1, Rna1, and Yrb1; Figure 5D and Figure 5—figure supplement 1A). tGFP-op (and NES-tGFP-op) had negative GIs with mutants in proteasome components but GFP-op did not (Figure 5D and Figure 5—figure supplement 1B). tGFP and NES-tGFP form aggregates, but not GFP (Figure 6C). This difference may be due to the higher molecular weight of tGFP compared to GFP and the presence of a repeating structure within the molecule. A larger molecule may increase the likelihood of misfolding during translation, or the presence of a repeating structure may increase intermolecular interactions and trigger the creation of large aggregates. Based on our results, we hypothesized a model explaining the negative genetic interaction between tGFP-op and the proteasome mutants (Figure 7—figure supplement 1). tGFP has a high probability of misfolding during translation, and when misfolded, it is ubiquitinated and degraded by the proteasome (1). In addition, tGFP also forms large intracellular aggregates that sequester proteasomes (2), ubiquitin (3), and chaperones (4). Among these, 1 will overload the proteasome's capacity (Kintaka et al., 2016). If 2 happens, it will lower the amount of proteasomes in the cell. If 3 happens, there would be a negative effect on proteostasis due to the depletion of ubiquitin (Higgins et al., 2020). The occurrence of 4 would have a similarly negative effect on proteostasis. In addition, since Ssa1/Ssa2 is required for proteasome assembly (Hammack et al., 2017), a depletion of these proteins would lower the amount of proteasomes. Based on the results of the proteome analysis, we believe that 4 is particularly likely, given that large amounts of Ssa1/Ssa2 were detected in the tGFP-op precipitates.

A comparison of mutants interacting with overproduction of three model proteins led to the isolation of mutants which specifically interact with GFP-op (Figure 5). The three model proteins caused growth defects with different expression levels (Figure 6B). The GFP level is considerably higher than the levels of tGFP and NES-tGFP, and its expression is the highest of all proteins in yeast (Eguchi et al., 2018), suggesting that overproduction of GFP causes growth defects because of the protein burden. As the protein burden should be triggered by the overproduction of otherwise non-harmful proteins like GFP (Moriya, 2015), these mutants should either exacerbate or mitigate the protein burden. The protein burden is considered to be growth defects occurring as a result of the overloading of protein synthesis processes (Kafri et al., 2016). In contrast to the expectation that mutants in those processes exacerbate the protein burden, the mutants isolated did not show any GO term enrichment in those processes but showed enrichment in actin-related processes like ‘cytoskeletal organization’ or ‘cellular bud’ (Figure 2D). Morphological analysis of cells also supported that GFP-op affects normal actin functions (Figure 7A–E). This relationship might be a result of the long-known connection between actin and translational machinery Kim and Coulombe, 2010; the protein burden-triggered growth defects might involve the perturbation of the actin cytoskeleton via translational factors like eEF1A, which can bundle actin fibers (Munshi et al., 2001). Mutations that mitigate the protein burden indeed enriched genes involved in protein synthesis, especially the transcriptional processes ‘RNA 3′-end processing’ and ‘RNA polymerase II transcription factor complex’ (Figure 2D). Because GFP expression levels in those mutants were lower than average (Figure 4D), those mutants might simply reduce the transcription of the GFP transcript itself.

It is thought that only harmless proteins can be produced up to ‘the ultimate expression level’ to cause the protein burden because harmful proteins should cause cellular defects at lower expression levels (Moriya, 2015). Those defects should be triggered by overloading more limited cellular resources, such as those used for folding and transport, accelerated non-specific interactions, or untimely activation of pathways (Moriya, 2015). Our study here supported this idea through the following observations: (1) tGFP (and NES-tGFP) consists of aggregates in the cell and thus could cause proteostasis stress (Figure 6C,D); (2) NES-tGFP further uses the protein export machinery; (3) genetic profiling suggested that tGFP-op and NES-tGFP-op overload the proteasome and protein export machinery (Figure 5D); (4) expression levels of tGFP and NES-tGFP, which cause growth defects are far lower than that of GFP (Figure 6B); and (5) GFP-op isolated specific mutants that were not isolated in tGFP-op and NES-tGFP-op. Figure 7F provides a schematic model summarizing this idea. Only harmless proteins like GFP can be produced up to the ultimate expression levels that cause the protein burden, which seems to be related to actin functions. Other proteins, localized or aggregative, can be produced at far lower levels than the level which causes the protein burden because their overproduction overloads localization or protein degradation resources which are more limited than the protein synthesis resource.

Finally, we propose a ‘barrel model’ to explain the relationship between the capacity of intracellular processes and the limits of expression (Figure 7G). In order for the cell to maintain its vital functions, resources within the cell are distributed to processes such as synthesis, folding, degradation, and transport. Each process has a fixed capacity depending on the amount of resources allocated to it (represented by the size of the barrel in Figure 7G). Each protein is synthesized, folded, transported, and degraded while using the resources in the cell (shown with arrows in Figure 7G). Overproduction of proteins processed by each process causes an overload of each resource, thus stalling the processing of other proteins produced by that process and creating cellular dysfunction. If a protein is processed by more than one process, its overexpression will first cause an overload of the process with the smallest capacity. Therefore, the level of overexpression of a protein that causes growth inhibition is likely to be determined by the process with the lowest capacity out of the processes by which the protein is processed. The capacity of synthesis, where all proteins are processed, should be the largest, and therefore proteins that undergo only the synthesis processing, that is those that fold on their own, localize to the cytoplasm, and do not undergo rapid degradation, are considered to have the highest expression limits (Such proteins are referred as gratuitous proteins here). Previous studie has suggested that GFP and Gpm1 are such proteins, and they cause growth inhibition when expressed up to 15% of total proteins (Kintaka et al., 2016; Eguchi et al., 2018). We define this proliferation inhibition effect as the protein burden. Protein burden is therefore thought to be caused by perturbations to protein synthesis. Our analysis of genetic interactions, however, did not provide a clear link to perturbations to the synthetic process; our results suggest that protein burden causes an unexpected perturbation of actin function. The details of this mechanism remain unclear at present, and further studies are needed. On the other hand, processes in which only a fraction of proteins are processed have a smaller capacity than synthesis, and the expression limits of proteins processed by those processes must be lower than the limits of protein burden. In fact, the expression limits of tGFP, which undergoes degradation by aggregation and ubiquitination (probably by misfolding), and NES-tGFP, which is transported outside the nucleus, were much lower than those of GFP. Since their expression limits were about 4% of GFP (Figure 6B), it can be estimated that their expression limits are about 0.6% of the total proteins. This amount may reflect the capacities of the processes by which these proteins are processed.

In conclusion, our genetic profiling successfully investigated the consequences of overproduction: overload of protein synthesis, nuclear export, and the proteasome. Mutants isolated in this study will be useful resources for further investigations into the general consequences of protein overproduction, especially the overloading of cellular processes.

Materials and methods

Strains and plasmids used in this study

Request a detailed protocol

The vector plasmid (pTOW40836), GFP-op plasmid (pTOW40836-TDH3pro-GFP), tGFP-op plasmid (pTOW40836-PYK1pro-tGFP), and NES-tGFP-op plasmid (pTOW40836-PYK1pro-NES-tGFP) have been described previously (Kintaka et al., 2016; Eguchi et al., 2018). Other than SGA, strains BY4741 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) and BY4743 (MATa/α his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 LYS2/lys2Δ0 met15Δ0/MET15 ura3Δ0/ura3Δ0) were used as wild-type strains in the analysis. The deletion mutant collection and temperature-sensitive mutant collection have been described previously (Costanzo et al., 2016). Yeast culture and transformation were performed as previously described (Amberg and Burke, 2005). A synthetic complete (SC) medium without uracil (Ura) or leucine (Leu) was used for yeast culture.

Query strains

Request a detailed protocol

Y7092 (MATa can1Δ::STE2pr-his5 lyp1Δ ura3Δ0 leu2Δ0 his3Δ1 met15Δ0) was used for the query train in the SGA. Y7092-E2-Crimson (MATa can1Δ::TDH3pr-E2-Crimson STE2pr-his5 lyp1Δ ura3Δ0 leu2Δ0 his3Δ1 met15Δ0) was used for the query strain in the SGA with the GFP fluorescent measurement experiment.

Synthetic genetic array (SGA) and colony size analysis

Request a detailed protocol

SGA and colony size analysis were performed as previously described (Baryshnikova et al., 2010). An empty plasmid (pTOW40836), and plasmids for overproducing GFP (pTOW40836-TDH3pro-GFP), tGFP (pTOW40836-PYK1pro-tGFP), and NES-tGFP (pTOW40836-PYK1pro-NES-tGFP) were introduced into the deletion and temperature-sensitive mutant collections using robots to manipulate libraries in 1536-colony high-density formats. A query strain harboring each of the overexproduction plasmids and each of the MATa mutant strains harboring a different genetic alteration were mated on YPD. Diploid cells were selected on plates containing both selection markers (YPD + G418 + clonNAT) found in the haploid parent strains. Sporulation was then induced by transferring cells to nitrogen starvation plates. Haploid cells containing all desired mutations were selected for by transferring cells to plates containing all selection markers (SC –His/Arg/Lys + canavanine + thialysine + G418 + cnonNAT) to select against remaining diploids. To analyze the growth of each deletion strain with the plasmids, all custom libraries were replicated to SC–LU plates and grown for three days at 30°C.

The fitness of each strain was assessed as normalized colony size on agar plates. Measurements of fitness and calculation of genetic interaction scores for each strain from colony images on agar plates were performed using SGA-tool (http://sgatools.ccbr.utoronto.ca) (Wagih et al., 2013). The colony size was quantified and normalized as shown in Figure 1—figure supplement 2. Then the genetic interaction (GI) scores were calculated using the formula. GI score (ε)=WAB – WA × WB, where WAB is overproduction-plasmid/mutant fitness, WA is Vecvtor control/mutant fitness, and WB is set to one as shown in Figure 1—figure supplement 2. The GI scores were filtered using the defined confidence threshold (GI score, |ε|>0.08), and p-value that reflects both the local variability of replicate colonies (four colonies/strain) and the variability of the strain sharing the same query or array mutation (p<0.05) (Baryshnikova et al., 2010). This filtered data set was used for all analyses.

For GFP-op_positive mutants, we further filtered the mutants as follows. Initial GFP-op_positive 146 genes (147 mutants) contained genes involved in the His and Lys synthetic pathways. His and Lys (Arg) are used as marker genes for the SGA, and deletion mutants of HIS, LYS, and ARG genes should not grow in the SGA analysis. In fact, the colony sizes of these mutants in the vector control experiment were very small and were considered to be the carryover. We thus further isolated positively-interacting mutants by setting a threshold on the colony size of greater than 0.39 in the vector control experiment, selected according to the largest colony size (ARG1) among the HIS, LYS, and ARG mutants, to avoid the identification of false-positive GIs.

Colony size mesurement

Request a detailed protocol

Colony size was measured by using the image analysis software SGA tools (http://sgatools.ccbr.utoronto.ca/) to determine accurate pixel colony sizes. Average values and standard deviations were calculated from at least six replications. Y7092 was used as the wild-type host strain.

Liquid growth measurement

Request a detailed protocol

Cellular growth was measured by monitoring OD595 every 30 min using a model 680 microplate reader (BioRad). The maximum growth rate was calculated as described previously (Moriya et al., 2006). Average values and standard deviations were calculated from biological triplicates. BY4741 was used as the wild type host strain.

GFP fluorescent measurement by typhoon

Request a detailed protocol

Two colonies/strain from the SGA were picked up and replicated to SC–U plates, and grown for two days at 30°C. To detect the fluorescence of the colony, plates were scanned by laser (GFP at 488 nm and E2-Crimson at 532 nm) using Typhoon 9210 (Amersham Biosciences). The image data were analyzed using GenePix Pro Software (Molecular Devices). Each colony was segmented by a circle with the same diameter, the fluorescence per pixel was detected, and the medians of the fluorescence in the circle were calculated. To normalize the intensity by plate, all medians were divided by the plate average median for GFP and E2-Crimson. The ratios of GFP/RFP were calculated, and the averages of the two colonies were used.

Clustering analysis

Request a detailed protocol

The GI scores of GFP, tGFP, and NES-tGFP were clustered into 15 clusters by the hierarchical clustering (average) method using R (https://www.r-project.org).

Enrichment analysis

Request a detailed protocol

Enrichment analysis was performed using the gene list tool on the Saccharomyces genome database (yeastmine.yeastgenome.org/yeastmine/bag.do) (Cherry et al., 2012).

Microscope observation

Request a detailed protocol

Log-phase cells were cultivated in SC–Ura medium. Cell images were obtained and processed using a DMI6000 B microscope and Leica Application Suite X software (Leica Microsystems). GFP fluorescence was observed using the GFP filter cube. Cellular DNA was stained with 100 μg/ml Hoechst 33342 (H3570, ThermoFisher) for 5 min and observed using the A filter cube. BY4741 was used as the host strain.

Quantification of GFP expression level

Request a detailed protocol

The total protein was extracted from log-phase BY4741 cells harboring overproduction plasmids with NuPAGE LDS sample buffer (ThermoFisher NP0007) after 0.2 N NaOH treatment for 5 min (Kushnirov, 2000). For each analysis, the total protein extracted from 0.1 optical density unit of cells OD600 1.0 was used. The extracted protein was labeled with Ezlabel FluoroNeo (WSE-7010, ATTO), as described in the manufacturer’s protocol, and separated by 4–12% SDS-PAGE. Proteins were detected and measured using a LAS-4000 image analyzer (GE Healthcare) in SYBR–green fluorescence detection mode, and Image Quant TL software (GE Healthcare). The intensity of the 45 kDa band corresponding to Pgk1 and Eno1/2 was used as the loading control. To detect GFP, the SDS-PAGE-separated proteins were transferred to a PVDF membrane (ThermoFisher). GFP was detected using an anti-GFP antibody (11814460001, Roche), a peroxidase-conjugated second antibody (414151F, Nichirei Biosciences), and a chemiluminescent reagent (34095, ThermoFisher). The chemiluminescent image was acquired with a LAS-4000 image analyzer in chemiluminescence detection mode (GE Healthcare). For the estimation of relative GFP levels, the intensities of corresponding GFP bands were normalized using the loading control described above.

Cell fractionation and detection of ubiquitination

Request a detailed protocol

BY4743 cells with an overproduction plasmid were cultured overnight at 30°C in 25 ml SC–Leu/Ura medium. Cells were collected, cells were suspended in 1 mL of lysis buffer 10 mM Phosphate buffered saline (pH7.4), 0.001% Tween20, Halt Protease Inhibitor Cocktail (78425, ThermoFisher). Glass beads were added to the cell suspension and the tube was vortexed three times for 2 min. The sample was chilled on ice for 3 min between vortexing. The cell lysate was centrifuged at 20,000 × g for 10 min, and the supernatant was transferred to another tube. The precipitates were washed five times with 1 mL of PBST. The final precipitates were suspended in 100 µL of PBST. The sample was treated with NuPAGE sample buffer (NP0007, ThermoFisher) at 70°C for 10 min., and the proteins were separated by SDS-PAGE. Total protein and GFP were detected by Ezlabel FluoroNeo and western blotting with GFP antibodies as described above. Detection of ubiquitin was performed by western blotting same as the one of GFP except that the anti-ubiquitin antibody (P4D1, Santa Cruz) was used.

Proteome analysis

Request a detailed protocol

Protein samples were precipitated by methanol/chloroform method and resolved in 0.1 M Tris-HCl (pH 8.5) containing 8 M urea. After reduction with DTT and alkylation with iodoacetoamide, urea concentrations were diluted by 4-fold with 0.1 M Tris-HCl (pH 8.5) and then digested into peptides by trypsin (Promega, Wisconsin, WI). Digested samples were centrifugated at 20,000 × g for 10 min and the tryptic peptides in supernatant were analysed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) system consisting of a DiNa nano LC (KYA technologies, Tokyo, Japan) and a LTQ-Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Acquired MS/Ms spectra were subjected to database search against protein sequences downloaded from the Saccharomyces Genome Database (http://www.yeastgenome.org/) by SEQUEST alogorithm. The number of peptide-spectrum match (PSM) for each protein, which fulfill the criteria of false discovery rate below 1%, was listed in Figure 6—source data 1.

High-dimensional morphological analysis

Request a detailed protocol

Morphological data of cells cultured were acquired as previously described (Ohya et al., 2005). Briefly, logarithmic-phase BY4741 cells harboring plasmids grown in SC–Ura medium were fixed and were triply stained with FITC-ConA, rhodamine-phalloidin, and 4,6-diamidino-2-phenylindole to obtain fluorescent images of the cell-surface mannoprotein, actin cytoskeleton, and nuclear DNA, respectively. Images of at least 200 individual cells were acquired and processed using CalMorph (version 1.2). All of the statistical analyses were performed with R. To statistically test the morphological differences among four strains, we conducted one-way ANOVA of the generalized linear model (GLM) for each of 501 morphological parameters. Probability density functions (PDFs) and accompanying link functions in the GLM were assigned to each trait as described previously (Yang et al., 2014). Difference of the four strains (n = 5) was incorporated as the explanatory variable into the linear model. We assessed a dispersion model among the strains in the linear models for the 501 parameters by Akaike information criterion (AIC) and set 110 models (parameters) as a different dispersion model because of lower AIC than that of a single dispersion model. Applying one-way ANOVA among the four strains to all 501 parameters, 51 of the 501 parameters were found to differ significantly at false discovery rate (FDR) = 0.01 by the likelihood ratio test (Likelihood ratio test in Figure 7—source data 1). Maximum likelihood estimation, likelihood ratio test, and the estimation of FDR were performed using the gamlss, lrtest, and qvalue functions in the gamlss (Stasinopoulos and Rigby, 2007), lmtest (Zeileis and Hothorn, 2002), and qvalue (Storey, 2002) R package. By Wald test at FDR = 0.01, 16, 17, and 24 of the 501 traits were detected to have a significant difference from wild-type in GFPy66g, Gpm1, and Gpm1-m, respectively (Q value of Wald test in Figure 7—source data 1). Of the 16 parameters detected in GFPy66g, 14 parameters were grouped into four independent morphological features by four principal components (explaining 60% of the variance) extracted from principal component analysis for the Z values of 109 replicates of his3Δ (Suzuki et al., 2018) as described previously (Ohnuki et al., 2012), and were used for the illustration of morphological features (Figure 7E, Morphological features in in Figure 7—source data 1).

References

  1. 1
    Methods in Yeast Genetics: A Cold Spring Harbor Laboratory Course Manual
    1. DC Amberg
    2. D Burke
    (2005)
    Cold Spring Harbor Laboratory Press.
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
    Overexpression of translation elongation factor 1A affects the organization and function of the actin cytoskeleton in yeast
    1. R Munshi
    2. KA Kandl
    3. A Carr-Schmid
    4. JL Whitacre
    5. AE Adams
    6. TG Kinzy
    (2001)
    Genetics 157:1425–1436.
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
    A direct approach to false discovery rates
    1. JD Storey
    (2002)
    Journal of the Royal Statistical Society: Series B 64:479–498.
    https://doi.org/10.1111/1467-9868.00346
  33. 33
  34. 34
  35. 35
  36. 36
    Diagnostic checking in regression relationships
    1. A Zeileis
    2. T Hothorn
    (2002)
    R News 2:7–10.

Decision letter

  1. Nir Ben-Tal
    Reviewing Editor; Tel Aviv University, Israel
  2. Patricia J Wittkopp
    Senior Editor; University of Michigan, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The study enhances our understanding of cellular effects of protein load.

Decision letter after peer review:

Thank you for submitting your article "Genetic Profiling of Protein Burden and Nuclear Export Overload" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Patricia Wittkopp as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

In "Genetic Profiling of Protein Burden and Nuclear Export Overload", Kintaka et al. performed a synthetic genetic array screen to explore the genetic interactions between mutants (DMA and TSA libraries) and protein burden. They used the 2 micron plasmid of Saccharomyces cerevisiae to explore the genetic interactions in response to three different protein burdens/overloads: (i) regular GFP; (ii) triple-GFP (tGFP), a much bigger protein that they found created some aggregates when highly expressed; and (iii) the nuclear export signal-containing tGFP (NES-tGFP). A similar study has already been published in this journal (Farkas et al., 2018), where the genetic interaction between the DMA library and yEVenus burden was explored. The present work, however, is much more extensive: it includes four biological and two experimental repeats, a second mutant library (TSA), different types of burden, and, most importantly, it checks the effect of higher burden levels using a stronger promoter.

Essential revisions:

Fundamental:

Overall, this work increases the knowledge about processes involved in protein stress alleviation. Yet, while the manuscript describes interesting phenomena, it does not provide follow-up findings to explain the main observations and therefore remains descriptive in nature. Here are concrete suggestions to this end.

1) In general, since there is no systematic follow-up on the main findings, it is difficult to appreciate how cells deal with the increased protein burden. In particular, as the authors noted, their findings are significantly different from other studies using similar reporter systems. Additional work is needed in order to substantiate the different pathways.

2) The authors overexpress GFP that does not have any physiological activity in yeast cells and therefore considered non-harmful. Yet, overexpression of tGFP shows genetic interaction with the proteasome. This may imply that tGFP undergoes post-transcriptional modification(s), such as ubiquitylation and this must be taken into consideration when attempting to explain the data. Accordingly, the authors need to examine the ubiquitylation status of their constructs, especially in proteasome where ubiquitin conjugates accumulate. Abolishing GFP ubiquitylation (using a lysineless variant, for example), may mitigate or abolish genetic interactions with proteasome subunits, suggesting a role for ubiquitin during overexpression.

3) tGFP-op levels are !~10 fold lower than GFP-op and yet only tGFP forms aggregates and affects cell viability in proteasome mutants. This is an interesting observation that requires some explanations. How tGFP-op is related to its tendency to aggregate and whether it is related to the proteasome activity should be verified before drawing a conclusion.

4) A main limitation of this work is the use of "tug of war" plasmid to generate the burden. This means that the plasmid copy number, and with it the GFP levels, will change across different mutations. The authors found, for example, that GFP levels are lower in many of the Mediator mutants. The reaction of the mutant cells to the burden should be measured not only by the growth effect but also by the burden level. The latter can be approximated by GPF levels or by the plasmid copy number. The authors report that in -Leu/Ura conditions, there are 30 copies of the plasmid. This statement is correct only for WT, and needs to be measured or estimated for the different mutations. This estimation, again, can be done by measuring the GFP levels of the different mutants. For the GFP set, they indeed used a method to measure the GFP levels. It would have helped to understand the mutants in Figure 3, if the GFP levels had been shown – this could easily be done in a supplementary figure. Fluorescence levels were, however, not done in the tGFP and NES-GFP experiments, or at least they are not being reported. It is important and probably critical to successful publication, that GFP levels be measured and reported, to help interpret the results correctly. At a minimum, this GFP levels should be given for those mutants above the GI threshold.

5) When the authors compared their results with those of Farkas et al., they didn't find any correlation. Neither study, incidentally, quantified GFP levels. Also, the present study used a plasmid with much higher expression levels, which will inevitably have obscured comparability. Farkas and co-workers used yEVenus and they showed that "yEVenus binds weakly, but significantly to certain molecular chaperones…". This binding, that may be unique to yEVenus, could also be part of the reason for the apparent differences in results. Farkas et al. showed that the burden effect was reduced by adding more AA to the media. Checking the AA effect on the GFP burden would have helped to reconcile whether the difference in results is a reflection of a different kind of burden, or whether it is mainly the results of a growth effect that wasn't normalized by GFP levels.

Presentation:

The writing of the manuscript and the interpretation of the data needs considerable expansion. The methods and the results are not described clearly. Please see major comments below.

1) Clarify the definition of genetic interactions.

1a) The term “genetic interactions” should be defined in the Introduction in a way that is specific to how it is used in this study, e.g. "In this study we screened for gene knockouts or knockdowns that had different impacts on growth depending on whether a green fluorescent protein was overexpressed."

1b) Clarify the meaning of positive and negative interactions, specifically whether the mutations are deleterious or beneficial. A positive interaction could either mean that GFPop alleviates a growth defect or enhances a growth advantage. Which one happens more? Probably the former but this needs to be clearer. This should be broken down in the figure, e.g. what fraction of orange dots signify a growth defect being alleviated vs. a growth advantage being enhanced? Can the authors use 4 colors, blue, light blue, orange, light orange? Since the goal here is to understand the biology of the cell and how GFPop changes it, these details seems important.

2) Clarify the protein burden.

2a) It is not a good idea to assume readers are familiar with previous publications using the tow system. Thus, most readers will assume that all mutant strains are expressing the same amount of GFP plus or minus some kind of noise (e.g. plasmid copy number variation due to unequal cell division). Please explain this in more detail in the Results section.

2b) Also, the term protein burden needs to be continuously explained. It is easy to interpret that term as meaning “the growth defect imposed by GFPop”. But it probably refers to the number of GFPop molecules produced. Is that right? Rather than using the term, “protein burden”, sometimes it would be OK to spell out what is meant, e.g. the number of GFPop molecules that burden the cell.

2c) An interesting question that arises is whether, in negative interactions, the expression of GFPop is enhancing the growth defect of the gene knockout/knockdown or the gene knockout is increasing the protein burden, e.g. increasing the level to which the tow system is able to express GFP. The authors have an impressive method of disentangling these two hypotheses, which they explain in Figure 4. But paragraph two of subsection “Investigation of GFP expression levels of mutants” of the paper, where these possibilities are enumerated, are unclear. One has to go back and work out many details including the possible types of positive and negative interactions, and how the tow system worked. A diagram or cartoon should be presented earlier in the paper, perhaps as Figure 1, which explains the logic showing the different possibilities that could be happening inside of cells and how the authors plan to disentangle them. The authors could depict 4 double-mutant cells, for example, one with a high level of GFP but a lower growth rate than either the GFPop strain or the mutant strain. Then they could explain in the legend the hypothesis as to what is happening inside this cell. Maybe a figure is unnecessary. But some more detailed explanation of how this system works and how the authors plan to us it to disentangle the possible things happening inside of cells should come up much earlier in the paper.

3) The growth regime is not clearly explained.

3a) The text in the first part of the Results section is very brief and relies heavily on Figure 1 to explain the experimental set up. Perhaps add a few more sentences to guide the reader. For example, the comment in paragraph two of subsection “Isolation of mutants that have genetic interactions with GFP-op” about “colonies” does not make sense. It should have already been stated that growth is measured in colonies and not in liquid culture. Otherwise the word “colonies” comes from nowhere.

3b) On that note, Figure 1C is confusing because these particular measurements were taken in liquid culture. Why are two different culturing methods used? Since 1C is meant as a control to show how the GFP affects growth when no knockout/knockdown is present, shouldn't this control be performed using the same method as the rest of the experiments?

4) Why were the particular GI threshold levels chosen? Why were the cutoffs chosen, including 0.08 in Figure 1 and 0.2, to define GFPop-positive and negative?

5) Why better correlation across conditions than within conditions? Figure 2A shows that even when taking only GI that exceed the 0.8 threshold, the correlation between replicates is less than 0.5 in the DMA -Ura condition. But Figure 2B shows that when comparing DMA -Ura to DMA -Ura/-Leu the correlation is greater than 0.5. How come the correlation between replicates is less than the correlation between these two different conditions? Is it because by averaging multiple replicates you get a better sense of the true GI value? This should be addressed.

6) Too many abbreviations. There are so many terms in this paper that are used to capture important concepts, e.g. positive interaction, protein burden, TMA, GFPunit_L, GFPunit_H, GFPop_positive, GFPop_negative. The reader loses track of how all of these things are related. The authors should not rely on these abbreviations so much and talk though these relationships, e.g. "Mutants with growth defects that were enhanced by GFP overexpression were also more likely to produce less GFP, indicating that the limit of GFP overproduction in these cells was lower than in other cells." Sentences like these would be so helpful in explaining what is actually going on.

7) Sometimes the Results section read a little bit like a list.

7a) This is a common issue in studies that report GO terms from different groups. Figure 3 is particularly list-like. The balance of the paper should shift away from listing GO categories and towards explaining and interpreting what is happening inside of cells as in paragraph two of subsection “Investigation of GFP expression levels of mutants”, and in Figure 6. For example, the authors were nicely able to do more with the actin and cell bud genes and show that indeed these cells had aberrant morphology. Also Figure 6F where the authors digest all of this information into a hypothesis about what is happening inside of cells was nice. Problematically, that hypothesis is unclear and will only become clear once points 1 and 2 above are addressed.

7b) Also, in 6F the relationship to the proteasome was not as clear as was the perturbation of actin. Could the proteasome relationship be explained more clearly?

Final thoughts: In sum, the authors' goal seems to be to go beyond listing GO categories to talk about how the protein burden affects cell biology. They need to rework the paper a bit in order to achieve this goal.

8) In the NES-tGFP experiment (and only in this experiment), they found strong interaction with the nucleus export machinery. This result isn't surprising per se but it's a good proof of concept, and it can be used as a control – showing that the system is working, which is worth mentioning. So this work could be an important resource to the field enabling a deeper understanding of protein burden origin. It should be emphasized.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Genetic Profiling of Protein Burden and Nuclear Export Overload" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Patricia Wittkopp as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

As you can see, reviewer 2, raised outstanding issues regarding the mechanism. Originally they felt that more experiments are needed. However, in the discussion that followed they agreed to allow the authors to deal with them by adding statements that further studies to reveal the mechanisms are needed. It is up to you to decide whether you would like to add more experiments or language changes.

Here is a part of the discussion that elaborate on the proposed experiments, which could be useful to you:

"One model that I can think of, which is based on recent literature, is that tGFP undergoes ubiquitylation and then a portion of the ubiquitylated protein is sequestered into large aggregates. Proteasomes are been recruited to these aggregate but the kinetics of degradation is altered. This results in overall reduced proteasome capacity, shown through genetic interaction with proteasome subunits. I proposed to the authors to address the connection between overexpression and aggregation.

In my second revision, I suggested two experiments:

1) to IP tGFP from the pellet and show that indeed it is the protein that undergo ubiquitylation. I believe that technically this experiment is essential.

2) To reduce overall ubiquitylation (through the overexpression of mutant R48 ubiquitin) and test if tGFP is still aggregated and if this has an effect on cell growth.

One can think of alternative experiments, but the point is that the basis to the effects the authors observed upon overexpression of tGFP is still unclear to me."

Otherwise, please re-phrase some of the statements regarding the ubiquitylation of tGFP and the possible mechanism of tGFP function. Regarding the former, showing poly-Ub chains in the pellet does not prove that tGFP is indeed as the substrate. Regarding the latter, according to the accumulating data, the mechanism is unlikely to be a general effect of overexpression that leads to the proteasome. A statement about the effect of aggregates on proteasomal degradation could also help clarifying the issue. Possibly the depletion of essential factors like molecular chaperones?

Reviewer #1:

The authors have addressed all of my previous comments, moreover, this is an impressive revision. The authors add new figures and text that clarify their methods and previous findings. The authors also add new analyses, including biochemical analyses of some of their overexpression strains. Finally, the authors add a new figure and synthesize their results into a model which describes the different ways cells might be affected by overexpressed proteins.

The revised paper represents a massive amount of experiments and careful thought about how the cell responds to overexpressed proteins. The paper sheds new light on this question.

Reviewer #2:

The authors nicely addressed my first comment. however, in my opinion, issues 2 and 3 require further clarification:

In respond to my comments regarding the level of expression and PTM of tGFP, the authors tested whether tGFP aggregates were ubiquitinated and concluded that overexpressed tGFP but not GFP forms ubiquitinated aggregates in cells. They hypothesized that tGFP-op causes an overload of the proteasome because tGFP is frequently misfolded, ubiquitinated and degraded by the proteasome. This may be the cause of the negative genetic interactions between tGFP-op and the proteasome mutants.

I suggest that this hypothesis should be further clarified, since aggregation of tGFP is at the center of the manuscript and without having a mechanistic explanation to its function, the significance of some the authors findings is unclear. Generally, the overexpression of misfolded proteins (even large ones) per se does not inhibit the proteasome in yeast cells. Alternatively, the recruitment of proteasomes to protein aggregates may abrogate their function. Since the proteasome harbor several ubiquitin receptors, it is possible that protesomes interact with aggregated proteins through conjugated ubiquitin chains. This could be tested for tGFP by isolating it from aggregates by IP and looking for ubiquitylation. Furthermore, I accept that having lysineless GFP might not be the best approach to tackle the issue. Yet, the authors could test the effect of conditional overexpression of lys48 ubiquitin mutant on aggregate formation, proteasome function and/or cell viability in cells overexpressing GFP or tGFP. This type of experiments should shade some light on the molecular basis for tGFP aggregation and the effect on cell growth.

Reviewer #3:

The revise version of "Genetic Profiling of Protein Burden and NuclearExport Overload", is greatly improved. the logic is easier to follow, and there are more illustrations. The comparison between their and Farkas' results are dipper, and most important to my opinion, they now have the GFP measurements for all the conditions, and those measurements now better integrated into the text and figures.

Their final model about the different ways protein overproduction can affect growth rate is very nice. they display some of the interesting biological questions that rise up from their screening, include the interaction between actin and protein burden, and that over-expression tGFP lead to aggregate and proteasome stress. They also did some follow-ups experiments to both of them, but they didn't reach to a biological understanding about the origin of those interesting observations: way Actin is so important to protein overproduction? and the reason for tGFP aggregates. It would have been nice to get a better understanding to at list one of them but many times interesting questions are as important as answers, and it is very extensive and interesting screen. So, they answered most of the question, and in my opinion the current wark important and good enough for publication.

https://doi.org/10.7554/eLife.54080.sa1

Author response

Essential revisions:

Fundamental:

Overall, this work increases the knowledge about processes involved in protein stress alleviation. Yet, while the manuscript describes interesting phenomena, it does not provide follow-up findings to explain the main observations and therefore remains descriptive in nature. Here are concrete suggestions to this end.

1) In general, since there is no systematic follow-up on the main findings, it is difficult to appreciate how cells deal with the increased protein burden. In particular, as the authors noted, their findings are significantly different from other studies using similar reporter systems. Additional work is needed in order to substantiate the different pathways.

We believe that the lack of reproducibility between our experimental results and those of Farkas et al. reflects the difficulty in obtaining genetic interactions when querying weak overexpression – protein burden. The rationale for this is as follows.

We found that reproducible genetic interactions are unlikely to be obtained from moderate overexpression of EGFP, as shown in Figure 2A and Figure 2—figure supplement 1. Moderate overexpression in our experiments was performed by expressing yEGFP from the TDH3 promoter cloned on a 2µ plasmid vector (Figure 2—figure supplement 3C). Similar to Farkas' conditions, overexpression of yEGFP from the HSC82 promoter cloned on a 2µ vector (Figure 2—figure supplement 3D) elicits milder growth inhibition under the moderate overexpression conditions. This suggests that the yEVenus overexpression condition used by Farkas et al. is milder than our -Ura condition and does not result in reproducible genetic interactions (reproducibility was not mentioned in Farkas et al., 2018 because Farkas et al. did not perform multiple experimental trials.)

Another basis for this is that the distribution of genetic interaction scores obtained by Farkas et al. is very narrow (see Figure 2—figure supplement 3A, B), and the threshold of the genetic interaction score for the presence of genetic interaction is 0.05. Our threshold is 0.2. If this threshold were used in the Farkas et al. analysis, only about 10 genes would show interaction.

Furthermore, we have confirmed the genetic interaction of GFP-op with the negative genes obtained by Farkas et al. and our analysis in individual experiments (Figure 2—figure supplement 4). While none of the seven negative genes obtained by Farkas et al. showed a negative genetic interaction with GFP-op, six of the seven negative genes we obtained reproduced the negative interaction with GFP-op.

Thus, we believe that this discrepancy reflects the difficulty in obtaining genetic interactions when a weak overexpression – protein burden is used as a query.

On the other hand, we cannot rule out the possibility that this result arises from differences in the experimental system we used, especially between the fluorescent proteins EGFP and Venus. What we want to know in this study, however, is not the physiological state that occurs when individual proteins such as EGFP and Venus are overexpressed, but the general physiological state of protein burden caused by the extreme overexpression of non-toxic gratuitous proteins, using EGFP and Venus as models. Therefore, there is no point in investigating genetic interactions (or vice versa) that can only be observed with either an excess of Venus or an excess of EGFP. We do not feel it makes sense to pursue this further in the present study, as it only investigates the differences in the characteristics of the two proteins.

In this regard, we have performed a multidimensional morphological analysis in the overexpressing strains, as shown in Figure 7. We show that actin mislocalization is caused not only by an excess of EGFP but also by an excess of another protein, Gpm1, indicating that perturbations to actin are more commonly caused by protein burden.

2) The authors overexpress GFP that does not have any physiological activity in yeast cells and therefore considered non-harmful. Yet, overexpression of tGFP shows genetic interaction with the proteasome. This may imply that tGFP undergoes post-transcriptional modification(s), such as ubiquitylation and this must be taken into consideration when attempting to explain the data. Accordingly, the authors need to examine the ubiquitylation status of their constructs, especially in proteasome where ubiquitin conjugates accumulate. Abolishing GFP ubiquitylation (using a lysineless variant, for example), may mitigate or abolish genetic interactions with proteasome subunits, suggesting a role for ubiquitin during overexpression.

We initially predicted that tGFP was just three GFPs bound together and that the effects of overexpression would be no different from those of GFP.

However, our experiments revealed that tGFP-op induces a distinctly different physiological state than GFP-op, because the expression limit of tGFP is only a few percent of that of GFP (original Figure 5F, new Figure 6B), tGFP-op and GFP-op genetically interact with different sets of genes (Figure 5C,D), and only tGFP forms giant aggregates in the cell (original Figure 5E, new Figure 6C).

While these findings are interesting, we did not see much point in pursuing this further, as tGFP is a rather specific, artificial protein with "three GFPs in a row".

On the other hand, in response to the reviewer's comments, we thought that pursuing these findings might reveal the properties of proteins that exhibit toxicity in a different way than protein burden. In particular, tGFP may be considered a model for high-molecular-weight proteins (75 kDa) or for proteins with a repetitive structure.

The significance of ubiquitination may be tested by analysis of EGFP without lysine, as pointed out by the reviewer. However, EGFP contains 20 lysines and it has been reported that their replacement with arginine induced misfolding (PMID: 22792305). Therefore, it is not appropriate to use lysineless EGFP to investigate the significance of ubiquitination, as overexpression of lysineless EGFP may cause unexpected results.

Instead, we used biochemical methods to test whether tGFP aggregates were ubiquitinated: GFP-op and tGFP-op cells were disrupted, separated into soluble and insoluble fractions, and then western blotted with anti-ubiquitin antibodies, respectively. The results showed a highly ubiquitinated high molecular weight ladder in the insoluble fraction of the tGFP-op strain (new Figure 6D). Western blotting with anti-GFP antibodies showed a high molecular weight ladder in the insoluble fraction of GFP-op as well as tGFP-op (new Figure 6D). However, the high-molecular-weight ladder found in the insoluble fraction of GFP-op was not detected by the anti-ubiquitin antibody. These results suggest that overexpressed tGFP but not GFP forms ubiquitinated aggregates in cells.

Our hypothesis is that tGFP-op causes an overload of the proteasome because tGFP is frequently misfolded, ubiquitinated and degraded by the proteasome. This may be the cause of the negative genetic interactions between tGFP-op and the proteasome mutants.

On the other hand, based on the detection of genetic interactions and ubiquitination, overexpression of single GFP did not cause ubiquitination or load on the proteasome. This may be related to the fact that tGFP is larger than GFP and has features such as a repeating structure of the same molecule strung together.

The above results have been added to new Figure 6D, and we created a new section on the analysis of tGFP.

3) tGFP-op levels are !~10 fold lower than GFP-op and yet only tGFP forms aggregates and affects cell viability in proteasome mutants. This is an interesting observation that requires some explanations. How tGFP-op is related to its tendency to aggregate and whether it is related to the proteasome activity should be verified before drawing a conclusion.

As described above, we performed a biochemical analysis of tGFP. We detected a high molecular weight ladder that was ubiquitinated in the insoluble fraction only in tGFP-op. This ladder is likely to be ubiquitinated tGFP because it is at a higher molecular weight than tGFP.

Although it is currently only a speculation, we believe that this difference is due to the higher molecular weight of tGFP compared to GFP and the presence of a repeating structure within the molecule. A larger molecule may increase the likelihood of misfolding during translation. The presence of a repeating structure may increase intermolecular interactions and trigger the creation of large aggregates. These abnormal proteins may be processed by degradation by the ubiquitin-proteasome system, thus creating a load on these pathways.

While we did write a short description of the expected nature of tGFP in the original manuscript in the Discussion, we have added a further statement above.

4) A main limitation of this work is the use of "tug of war" plasmid to generate the burden. This means that the plasmid copy number, and with it the GFP levels, will change across different mutations. The authors found, for example, that GFP levels are lower in many of the Mediator mutants. The reaction of the mutant cells to the burden should be measured not only by the growth effect but also by the burden level. The latter can be approximated by GPF levels or by the plasmid copy number. The authors report that in -Leu/Ura conditions, there are 30 copies of the plasmid. This statement is correct only for WT, and needs to be measured or estimated for the different mutations. This estimation, again, can be done by measuring the GFP levels of the different mutants. For the GFP set, they indeed used a method to measure the GFP levels. It would have helped to understand the mutants in Figure 3, if the GFP levels had been shown – this could easily be done in a supplementary figure. Fluorescence levels were, however, not done in the tGFP and NES-GFP experiments, or at least they are not being reported. It is important and probably critical to successful publication, that GFP levels be measured and reported, to help interpret the results correctly. At a minimum, this GFP levels should be given for those mutants above the GI threshold.

While we agree that there are limitations to overexpression using the gTOW plasmid, these limitations do not only exist for overexpression using the gTOW plasmid, but are potentially present in any overexpression experiment.

In the case of overexpression using 2µ plasmids, not only gTOW plasmids, the effect of mutants on plasmid copy number is inevitable. In fact, as shown in Figure 2—figure supplement 5, retention of the multicopy plasmid itself can negatively affect growth in replication mutants. A similar hidden bias may be present in the experiments of Farkas et al. that preceded the present study, as they used 2µ plasmids.

On the other hand, limiting is also present in promoter substitution, which is a possible alternative to overexpression using multicopy plasmids. The most commonly used overexpression system in yeast is the GAL1 promoter, but overexpression of EGFP by the GAL1 promoter from a single copy cannot cause growth inhibition, i.e., it cannot be overexpressed to the point of causing protein burden. Furthermore, even if SGA analysis is performed using overexpression by GAL1 promoter substitution as a query, it is not possible to avoid the effects of mutants that alter the expression levels from the GAL1 promoter. For example, positive interactions could be falsely detected in mutants of the basic transcription factor, such as those obtained in this study, because of the reduced degree of overexpression.

In other words, no matter what overexpression system is used, the risk that mutations affecting the system will be obtained as pseudo-positive or negative interactions cannot be ruled out.

As noted by the reviewer, we used GFP expression levels to address these issues (as discussed below, this principle is explained in the new Figure 4—figure supplement 1). In response to the reviewer's comments, we have added the GFP expression data to Figure 3. We also analyzed the expression levels of tGFP-op and NES-tGFP as well as GFP-op. These results supported that the nuclear transport mutants are sensitive to NES-tGFP overexpression. (New Figure 5E). It was also predicted that either the expression of tGFP is elevated in proteasome mutants or that those mutants are sensitive to tGFP overexpression (Figure 6A).

We have added these results as a new figure (Figure 5E, Figure 5—figure supplement 2, Figure 6A) and added the relevant description to Results.

5) When the authors compared their results with those of Farkas et al., they didn't find any correlation. Neither study, incidentally, quantified GFP levels. Also, the present study used a plasmid with much higher expression levels, which will inevitably have obscured comparability. Farkas and co-workers used yEVenus and they showed that "yEVenus binds weakly, but significantly to certain molecular chaperones…". This binding, that may be unique to yEVenus, could also be part of the reason for the apparent differences in results. Farkas et al. showed that the burden effect was reduced by adding more AA to the media. Checking the AA effect on the GFP burden would have helped to reconcile whether the difference in results is a reflection of a different kind of burden, or whether it is mainly the results of a growth effect that wasn't normalized by GFP levels.

As shown in Figure 2—figure supplement 3, we and Farkas both used a 2µ plasmid. Our gTOW plasmid showed a higher copy number only under the -Ura/Leu condition, and we believe that under the -Ura condition the copy number is similar to that of the normal 2µ plasmid and therefore the same as that of the YEplac181 plasmid used by Farkas and co-workers. Since Farkas et al. use a promoter for HSC82, it is likely to be weaker than expression from our TDH3 promoter, but the degree of growth inhibition is not significantly different (Figure 2—figure supplement 3E). In other words, we believe that EGFP and Venus are expressed under our low-copy conditions at levels similar to the overexpression conditions of Farkas et al.

And, as shown in Figure 2—figure supplement 3, our experimental results are not consistent with those of Farkas et al. under these conditions as well. However, we believe that this lack of reproducibility reflects the difficulty in reproducibly obtaining genetic interactions under conditions of weak overexpression (Figure 2—figure supplement 1), as we already noted in section 1.

We adopted the high-copy condition for the present study because of its high reproducibility. The statistical analysis of the confident gene set showing genetic interactions under high-copy conditions did not suggest that GFP-op adversely affects HSP70-related proteostasis. Therefore, perturbations to HSP70-associated proteostasis should not be considered a general consequence of protein burden, or at least not a major consequence, and we do not see the significance of following up previous studies that follow that assumption in the present study.

The protein burden, as we see it, is a phenomenon caused by a "general" excess of a protein, not by a specific function of the overexpressed protein. If yEVenus binds to a specific chaperone and so depletes it, but EGFP does not because of its lack of binding to that chaperone, then we are not looking at a general phenomenon of protein burden. In fact, we are currently working on a study to evaluate the differences in cytotoxicity of EGFP, Venus, and other fluorescent proteins and will submit a paper soon, but this is beyond the scope of this paper and will not be mentioned in this paper.

On the other hand, since the burden effect on amino acid starvation can be immediately verified in our experimental setup, we performed the experiments. We measured the cost of GFP-op and yEVenus-op at diluted amino acid concentrations, as required by the reviewers (Figure 2—figure supplement 3G).

The results of Farkas et al. that the cost of yEVenus overexpression increased with decreasing amino acid concentration of the medium were confirmed in our experimental system. The cost of GFP overexpression also increased with decreasing amino acid concentration of the medium, but this increase was rather more pronounced than the increase in the cost of yEVenus. These results suggest that overexpression of GFP may impose a similar burden to that of yEVenus overexpression. On the other hand, the results also suggest that the effects of the excess of the two fluorescent proteins on cells are not entirely the same.

These results have been added to Figure 2—figure supplement 3G and the relevant description has been added to the Results.

Presentation:

The writing of the manuscript and the interpretation of the data needs considerable expansion. The methods and the results are not described clearly. Please see major comments below.

1) Clarify the definition of genetic interactions.

1a) The term “genetic interactions” should be defined in the Introduction in a way that is specific to how it is used in this study, e.g. "In this study we screened for gene knockouts or knockdowns that had different impacts on growth depending on whether a green fluorescent protein was overexpressed."

Following the reviewer's advice, We have added the following text to the Introduction.

“To understand the physiological conditions caused by protein burden, we conducted a systematic survey of mutants that exacerbate or alleviate the growth inhibition caused by GFP overexpression. We surveyed genetic interactions between mutant strains and high levels of GFP overproduction (GFP-op) to genetically profile cells exhibiting this phenomenon. Here, if a mutation exacerbates growth inhibition by GFP-op, or if GFP-op exacerbates growth inhibition by the mutation, the mutation has a negative genetic interaction with GFP-op. Also, if a mutation alleviates growth inhibition caused by GFP-op, the mutation has a positive genetic interaction with GFP-op. If GFP-op relaxes the growth inhibition caused by the mutation, it is also detected as a positive genetic interaction.”

1b) Clarify the meaning of positive and negative interactions, specifically whether the mutations are deleterious or beneficial. A positive interaction could either mean that GFPop alleviates a growth defect or enhances a growth advantage. Which one happens more? Probably the former but this needs to be clearer. This should be broken down in the figure, e.g. what fraction of orange dots signify a growth defect being alleviated vs. a growth advantage being enhanced? Can the authors use 4 colors, blue, light blue, orange, light orange? Since the goal here is to understand the biology of the cell and how GFPop changes it, these details seems important.

The interpretation of the genetic interaction was added in the Introduction section as described above.

We did not anticipate this reviewer's perspective on positive genetic interactions. We appreciate your point of view.

We assumed that the background principle behind the "positive genetic interaction" would be that mutations mitigate the growth-inhibiting effects of GFP-op (Case 1). For example, as our analysis suggested, mutants of the basic transcription factor appear to lower the expression of GFP, thus lowering the growth inhibitory effect of GFP-op. Also, as we recently published in our paper, mutations in chromatin remodeling factors can alleviate the growth inhibition by GFP-op by lowering the transcription of a specific group of genes to create more resource space (Saeki et al., 2020).

On the other hand, it is certainly possible that GFP overexpression could alleviate the growth inhibition caused by the mutation (Case 2). Furthermore, if GFP-op further enhances the growth of strains whose growth is improved by the mutation (Case 3), this would be detected as a positive genetic interaction. With regard to Case 1 and 2, however, we cannot distinguish between them, as the present analysis only provides a value of "GFP-op and the mutation's growth-inhibiting effect is less than expected". With respect to Case 3, we can observe this as a phenomenon where "the growth of the GFP-op im the mutant is higher than the growth of the wild type with the vector".

However, as we will discuss later, our screening does not allow us to directly compare fitness between plates, as we correct for fitness within plates. Since GFP-op and vector controls are analyzed on separate plates, colony sizes on each plate cannot be directly compared. On the other hand, it would be very interesting to know if there are mutants in GFP-op that have a higher growth rate than the wild type (+vector control). Therefore, we investigated the mutant strains that showed a positive genetic interaction with GFP-op and a higher fitness in the vector control than in the wild type. Such mutants can be selected computationally by taking into account the inhibition of growth by GFP-op. By this revision, we assessed the fitness reduction by GFP-op, tGFP-op and NES-tGFP-op on the same plate (revised Figure 5B). The fitness reduction by GFP-op obtained there (0.78) was used to correct for colony size on GFP-op. The results of such a screening are shown in new Figure 4—figure supplement 3. As a result, 18 mutants were obtained. In addition, 14 mutants were obtained when the strains with higher than average GFPunit were screened. These were subjected to enrichment analysis and three components of the microtubule-binding protein DAM/DASH complex were found to be contained in these 14 mutants. At present, the molecular mechanism for why these mutants behave in this way is not clear. Individual analyses of these strains will be necessary in the future. This data is shown in new Figure 4—figure supplement 3 and 4D, and the relevant description has been added to Results.

In the process of this analysis, we found some errors in the graph in Figure 2D, which we have replaced. Some positive genes that were excluded from the analysis had been shown as orange circles in the original Figure 2D.

2) Clarify the protein burden.

2a) It is not a good idea to assume readers are familiar with previous publications using the tow system. Thus, most readers will assume that all mutant strains are expressing the same amount of GFP plus or minus some kind of noise (e.g. plasmid copy number variation due to unequal cell division). Please explain this in more detail in the Results section.

In response to the reviewer's comments, we have added a detailed description of the gTOW system to new Figure 1—figure supplement 1. We have also added a detailed description of the gTOW system to Figure 4—figure supplement 1 to show how copy number variation by gTOW affects GFP expression levels.

2b) Also, the term protein burden needs to be continuously explained. It is easy to interpret that term as meaning “the growth defect imposed by GFPop”. But it probably refers to the number of GFPop molecules produced. Is that right? Rather than using the term, “protein burden”, sometimes it would be OK to spell out what is meant, e.g. the number of GFPop molecules that burden the cell.

Thanks for the advice. We define protein burden as the phenomenon of growth defects that occurs when a gratuitous protein is extremely overexpressed in the cell. Indeed, there was some ambiguity in the use of protein burden in the Results. We have changed the wording in this section to clarify that the protein burden is a growth inhibition phenomenon.

2c) An interesting question that arises is whether, in negative interactions, the expression of GFPop is enhancing the growth defect of the gene knockout/knockdown or the gene knockout is increasing the protein burden, e.g. increasing the level to which the tow system is able to express GFP. The authors have an impressive method of disentangling these two hypotheses, which they explain in Figure 4. But paragraph two of subsection “Investigation of GFP expression levels of mutants” of the paper, where these possibilities are enumerated, are unclear. One has to go back and work out many details including the possible types of positive and negative interactions, and how the tow system worked. A diagram or cartoon should be presented earlier in the paper, perhaps as Figure 1, which explains the logic showing the different possibilities that could be happening inside of cells and how the authors plan to disentangle them. The authors could depict 4 double-mutant cells, for example, one with a high level of GFP but a lower growth rate than either the GFPop strain or the mutant strain. Then they could explain in the legend the hypothesis as to what is happening inside this cell. Maybe a figure is unnecessary. But some more detailed explanation of how this system works and how the authors plan to us it to disentangle the possible things happening inside of cells should come up much earlier in the paper.

We agree that the original manuscript did not explain this part well enough and it was difficult for the reader to understand it. Therefore, we have added an interpretation of this experiment as new Figure 4B. We have also created a new figure (Figure 4—figure supplement 1) to illustrate the detailed background principles.

As described above, we also analyzed the GFP fluorescence for tGFP and NES-tGFP. This increased the number of concrete examples of gene sets belonging to the four categories, and we hope that the readers can now get a clearer picture of what these four categories mean.

The reviewer suggests that this analysis of GFP fluorescence and its interpretation be explained earlier in the Results. However, as the reviewer thinks, this analysis is a bit complicated and not easy to understand, so we would like to maintain the current order of explanation, first acquiring mutant strains by genetic interactions and then classifying them by GFP expression levels, in order to allow the readers to trace our thoughts.

3) The growth regime is not clearly explained.

3a) The text in the first part of the Results section is very brief and relies heavily on Figure 1 to explain the experimental set up. Perhaps add a few more sentences to guide the reader. For example, the comment in paragraph two of subsection “Isolation of mutants that have genetic interactions with GFP-op” about “colonies” does not make sense. It should have already been stated that growth is measured in colonies and not in liquid culture. Otherwise the word “colonies” comes from nowhere.

We apologize for our lack of explanation. Following the reviewer's suggestion, we have added a text explaining how we measured growth (fitness) to the Results.

There is an established pipeline for calculating genetic interactions using the synthetic genetic array (e.g. Costanzo et al., 2016) and we have calculated genetic interactions accordingly.

On the other hand, because this analysis differs in some respects from the usual genetic interaction analysis, we have created a new figure (Figure 1—figure supplement 2) to explain the calculation of genetic interactions in this study. In the process of preparing this figure and explanation, we found some inaccuracies in the methods of the original manuscript, so we corrected them.

3b) On that note, Figure 1C is confusing because these particular measurements were taken in liquid culture. Why are two different culturing methods used? Since 1C is meant as a control to show how the GFP affects growth when no knockout/knockdown is present, shouldn't this control be performed using the same method as the rest of the experiments?

I agree with the reviewer's comment. We have measured growth at colony size and added it to a new figure (Figure 1C, D). We also performed colony-sized growth tests for tGFP-op and NES-tGFP-op and replaced Figure 5B with a new one, and changed the description according to the data.

4) Why were the particular GI threshold levels chosen? Why were the cutoffs chosen, including 0.08 in Figure 1 and 0.2, to define GFPop-positive and negative?

We chose these thresholds based on the previous publications, and to isolate confident (and strong) genetic interactions. We added the rationale to Results.

5) Why better correlation across conditions than within conditions? Figure 2A shows that even when taking only GI that exceed the 0.8 threshold, the correlation between replicates is less than 0.5 in the DMA -Ura condition. But Figure 2B shows that when comparing DMA -Ura to DMA -Ura/-Leu the correlation is greater than 0.5. How come the correlation between replicates is less than the correlation between these two different conditions? Is it because by averaging multiple replicates you get a better sense of the true GI value? This should be addressed.

We were simply showing this data to show that our mutant isolation method is effective.

Since the reproducibility between replicates at the 0.08 threshold for TSA is 0.70 (Figure 2A, TSA-0.08) and between conditions is 058 (Figure 2C), the reproducibility between replicates is higher in the case of TSA. On the other hand, indeed, for the DMA, the correlation is higher between conditions (0.70) than between replicates (0.62).

The reason for this is unclear, but as the reviewer points out, it may be an effect of averaging. We have therefore added the following statement to our Results:

“We note that there is a higher correlation between conditions at -Ura and -Leu/Ura than between replicates in DMA (Figure 2A and Figure 2B). The cause of this is unclear, but it may indicate that averaging between replicates yields values closer to the true GI score.”

6) Too many abbreviations. There are so many terms in this paper that are used to capture important concepts, e.g. positive interaction, protein burden, TMA, GFPunit_L, GFPunit_H, GFPop_positive, GFPop_negative. The reader loses track of how all of these things are related. The authors should not rely on these abbreviations so much and talk though these relationships, e.g. "Mutants with growth defects that were enhanced by GFP overexpression were also more likely to produce less GFP, indicating that the limit of GFP overproduction in these cells was lower than in other cells." Sentences like these would be so helpful in explaining what is actually going on.

Perhaps the abbreviation of GFPunit and its concept are the most confusing. We have added a new table of this as Figure 4B, as described above. We hope to avoid some confusion by referring the reader to this.

For the DMA and TSA in the figure, we have rewritten them to spell them out as much as possible.

7) Sometimes the Results section read a little bit like a list.

7a) This is a common issue in studies that report GO terms from different groups. Figure 3 is particularly list-like. The balance of the paper should shift away from listing GO categories and towards explaining and interpreting what is happening inside of cells as in paragraph two of subsection “Investigation of GFP expression levels of mutants”, and in Figure 6. For example, the authors were nicely able to do more with the actin and cell bud genes and show that indeed these cells had aberrant morphology. Also Figure 6F where the authors digest all of this information into a hypothesis about what is happening inside of cells was nice. Problematically, that hypothesis is unclear and will only become clear once points 1 and 2 above are addressed.

The first half of this study is an attempt to identify the biological function of the genes we have isolated, using only statistical and objective indicators. We believe that enrichment analysis using GO (and publications) is currently the only way to objectively demonstrate the biological significance of a data set. The results of enrichment analysis must currently be a list. We feel, too, that there is certainly a need for a method of presenting the results of GO analysis in a graphical way that appeals to people's intuition.

On the other hand, we believe that Figure 3 is more than just a list of enrichment analysis results. Figure 3 allows the readers to know what exactly the names of the genes in each category of GO are. It will give more information than just the categories, especially to the experts. We also hope that by looking at this graph, the reader can see that there are multiple alleles in the temperature-sensitive strains, and that these alleles uniformly tend to show a positive/negative genetic interaction. Furthermore, we believe that the reader will understand that the gene cluster as a whole tends to exhibit positive/negative interactions beyond the noise of the experimental system. In addition to that, we have also added GFP expression levels to this graph in this revision. This graph should give the readers a clearer picture of the analysis performed in this experiment.

7b) Also, in 6F the relationship to the proteasome was not as clear as was the perturbation of actin. Could the proteasome relationship be explained more clearly?

Final thoughts: In sum, the authors' goal seems to be to go beyond listing GO categories to talk about how the protein burden affects cell biology. They need to rework the paper a bit in order to achieve this goal.

That's right. We are only using enrichment analysis to objectively arrive at biological information from our data set.

In this revision, we analyzed GFP expression levels for tGFP and NES-tGFP, and we also performed biochemical analysis of tGFP (and GFP) overexpression strains. We hope that by presenting these analyses, we have clarified our original goal of understanding the physiological state of cells caused by overexpression by analyzing genetic interactions.

We have added this result as new Figure 5 and Figure 6 and have added relevant descriptions to the results.

We have further added a model as Figure 7G that clarifies the conceptualization of this study and a related description to Discussion.

8) In the NES-tGFP experiment (and only in this experiment), they found strong interaction with the nucleus export machinery. This result isn't surprising per se but it's a good proof of concept, and it can be used as a control – showing that the system is working, which is worth mentioning. So this work could be an important resource to the field enabling a deeper understanding of protein burden origin. It should be emphasized.

Thank you for the suggestion. Indeed, we also think this is the clearest result of the proof of concept. We have now also analyzed the data for GFP expression in NES-tGFP to show the results more clearly as Figure 5E. We have also rewritten the text to strengthen our argument.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

As you can see, reviewer 2, raised outstanding issues regarding the mechanism. Originally they felt that more experiments are needed. However, in the discussion that followed they agreed to allow the authors to deal with them by adding statements that further studies to reveal the mechanisms are needed. It is up to you to decide whether you would like to add more experiments or language changes.

Here is a part of the discussion that elaborate on the proposed experiments, which could be useful to you:

"One model that I can think of, which is based on recent literature, is that tGFP undergoes ubiquitylation and then a portion of the ubiquitylated protein is sequestered into large aggregates. Proteasomes are been recruited to these aggregate but the kinetics of degradation is altered. This results in overall reduced proteasome capacity, shown through genetic interaction with proteasome subunits. I proposed to the authors to address the connection between overexpression and aggregation.

In my second revision, I suggested two experiments:

1) to IP tGFP from the pellet and show that indeed it is the protein that undergo ubiquitylation. I believe that technically this experiment is essential.

2) To reduce overall ubiquitylation (through the overexpression of mutant R48 ubiquitin) and test if tGFP is still aggregated and if this has an effect on cell growth.

One can think of alternative experiments, but the point is that the basis to the effects the authors observed upon overexpression of tGFP is still unclear to me."

Otherwise, please re-phrase some of the statements regarding the ubiquitylation of tGFP and the possible mechanism of tGFP function. Regarding the former, showing poly-Ub chains in the pellet does not prove that tGFP is indeed as the substrate. Regarding the latter, according to the accumulating data, the mechanism is unlikely to be a general effect of overexpression that leads to the proteasome. A statement about the effect of aggregates on proteasomal degradation could also help clarifying the issue. Possibly the depletion of essential factors like molecular chaperones?

We have made two revisions to answer this comment. We performed one experiment to identify proteins in the precipitated fraction of cells of tGFP-op and added a discussion of some mechanisms that might explain the negative genetic interaction between tGFP-op and the proteasome mutant predicted by the experimental results. We believe that this has allowed us to propose a clearer hypothesis about the mechanism of the negative genetic interaction between tGFP-op and the proteasome mutants.

Instead of performing the immunoprecipitation experiments suggested by the reviewer, we decided to perform a more direct experiment, i.e., systematic identification by LC-MS/MS of proteins in a precipitation fraction containing a large amount of aggregates generated by tGFP-op. We have added the results to the new Figure 6E and relevant description in Results. Instead, due to space constraints, we moved part of the older Figure 6D to Figure 6—figure supplement 1C.

As a result, proteasomes have been detected in the precipitation. Their amount was higher in tGFP-op than in the vector control, but it was still very small, so it was difficult to conclude that they were trapped in the aggregates and their intracellular concentration was reduced. On the other hand, Ssa1/Ssa2 chaperones, which are Hsp70, were detected in large quantities in the precipitation fraction of tGFP-op, as the reviewer had expected. It has been reported that Ssa1/Ssa2 is required for proteasome assembly, and we therefore believe that this result provides a promising model to explain the negative genetic interaction between tGFP-op and the proteasome mutants.

However, we believe that further experiments are needed to draw conclusions about the above as well. Therefore, as another revise, we have added our model for explaining the genetic interaction between tGFP-op and the proteasome mutant based on the experimental results as a new Figure 7—figure supplement 1, and added this explanation to the Discussion.

Reviewer #2:

The authors nicely addressed my first comment. however, in my opinion, issues 2 and 3 require further clarification:

In respond to my comments regarding the level of expression and PTM of tGFP, the authors tested whether tGFP aggregates were ubiquitinated and concluded that overexpressed tGFP but not GFP forms ubiquitinated aggregates in cells. They hypothesized that tGFP-op causes an overload of the proteasome because tGFP is frequently misfolded, ubiquitinated and degraded by the proteasome. This may be the cause of the negative genetic interactions between tGFP-op and the proteasome mutants.

I suggest that this hypothesis should be further clarified, since aggregation of tGFP is at the center of the manuscript and without having a mechanistic explanation to its function, the significance of some the authors findings is unclear. Generally, the overexpression of misfolded proteins (even large ones) per se does not inhibit the proteasome in yeast cells. Alternatively, the recruitment of proteasomes to protein aggregates may abrogate their function. Since the proteasome harbor several ubiquitin receptors, it is possible that protesomes interact with aggregated proteins through conjugated ubiquitin chains. This could be tested for tGFP by isolating it from aggregates by IP and looking for ubiquitylation. Furthermore, I accept that having lysineless GFP might not be the best approach to tackle the issue. Yet, the authors could test the effect of conditional overexpression of lys48 ubiquitin mutant on aggregate formation, proteasome function and/or cell viability in cells overexpressing GFP or tGFP. This type of experiments should shade some light on the molecular basis for tGFP aggregation and the effect on cell growth.

To answer the concern of reviewer #2, we have performed two revisions as described above.

https://doi.org/10.7554/eLife.54080.sa2

Article and author information

Author details

  1. Reiko Kintaka

    Donnelly Center for Cellular and Biomolecular Research, Department of Medical Genetics, University of Toronto, Toronto, Canada
    Contribution
    Formal analysis, Investigation, Visualization, Writing - review and editing
    Competing interests
    No competing interests declared
  2. Koji Makanae

    Research Core for Interdisciplinary Sciences, Okayama University, Okayama, Japan
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8560-4561
  3. Shotaro Namba

    Matching Program Course, Okayama University, Okayama, Japan
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Hisaaki Kato

    Graduate School of Environmental and Life Science, Okayama University, Okayama, Japan
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  5. Keiji Kito

    Department of Life Sciences, School of Agriculture, Meiji University, Tokyo, Japan
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9057-1688
  6. Shinsuke Ohnuki

    Graduate School of Frontier Sciences, University of Tokyo, Tokyo, Japan
    Contribution
    Formal analysis, Visualization, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Yoshikazu Ohya

    Graduate School of Frontier Sciences, University of Tokyo, Tokyo, Japan
    Contribution
    Resources, Supervision, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0837-1239
  8. Brenda J Andrews

    Donnelly Center for Cellular and Biomolecular Research, Department of Medical Genetics, University of Toronto, Toronto, Canada
    Contribution
    Resources, Supervision, Funding acquisition, Project administration
    Competing interests
    No competing interests declared
  9. Charles Boone

    1. Donnelly Center for Cellular and Biomolecular Research, Department of Medical Genetics, University of Toronto, Toronto, Canada
    2. RIKEN Center for Sustainable Resource Science, Wako, Japan
    Contribution
    Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  10. Hisao Moriya

    1. Research Core for Interdisciplinary Sciences, Okayama University, Okayama, Japan
    2. Graduate School of Environmental and Life Science, Okayama University, Okayama, Japan
    Contribution
    Conceptualization, Supervision, Funding acquisition, Investigation, Visualization, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    hisaom@cc.okayama-u.ac.jp
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7638-3640

Funding

Japan Society for the Promotion of Science (17H03618)

  • Hisao Moriya

Japan Society for the Promotion of Science (15KK0258)

  • Hisao Moriya

Japan Society for the Promotion of Science (18K19300)

  • Hisao Moriya

Japan Society for the Promotion of Science (20H03242)

  • Hisao Moriya

Japan Society for the Promotion of Science (19H03205)

  • Yoshikazu Ohya

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank members of the Moriya lab, the Boone lab and the Andrews lab for advice and helpful discussions. This work was partly supported by JSPS KAKENHI grant numbers 15KK0258, 17H03618, 18K19300, 20H03242.

Senior Editor

  1. Patricia J Wittkopp, University of Michigan, United States

Reviewing Editor

  1. Nir Ben-Tal, Tel Aviv University, Israel

Publication history

  1. Received: December 2, 2019
  2. Accepted: November 1, 2020
  3. Accepted Manuscript published: November 4, 2020 (version 1)
  4. Version of Record published: November 18, 2020 (version 2)

Copyright

© 2020, Kintaka et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,318
    Page views
  • 143
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Chromosomes and Gene Expression
    2. Genetics and Genomics
    Konstantin Riege et al.
    Tools and Resources

    The transcription factor p53 is the best-known tumor suppressor, but its sibling p63 is a master regulator of epidermis development and a key oncogenic driver in squamous cell carcinomas (SCC). Despite multiple gene expression studies becoming available, the limited overlap of reported p63-dependent genes has made it difficult to decipher the p63 gene regulatory network. Particularly, analyses of p63 response elements differed substantially among the studies. To address this intricate data situation, we provide an integrated resource that enables assessing the p63-dependent regulation of any human gene of interest. We use a novel iterative de novo motif search approach in conjunction with extensive ChIP-seq data to achieve a precise global distinction between p53 and p63 binding sites, recognition motifs, and potential co-factors. We integrate these data with enhancer:gene associations to predict p63 target genes and identify those that are commonly de-regulated in SCC representing candidates for prognosis and therapeutic interventions.

    1. Chromosomes and Gene Expression
    2. Genetics and Genomics
    Nour J Abdulhay et al.
    Tools and Resources

    Our understanding of the beads-on-a-string arrangement of nucleosomes has been built largely on high-resolution sequence-agnostic imaging methods and sequence-resolved bulk biochemical techniques. To bridge the divide between these approaches, we present the single-molecule adenine methylated oligonucleosome sequencing assay (SAMOSA). SAMOSA is a high-throughput single-molecule sequencing method that combines adenine methyltransferase footprinting and single-molecule real-time DNA sequencing to natively and nondestructively measure nucleosome positions on individual chromatin fibres. SAMOSA data allows unbiased classification of single-molecular 'states' of nucleosome occupancy on individual chromatin fibres. We leverage this to estimate nucleosome regularity and spacing on single chromatin fibres genome-wide, at predicted transcription factor binding motifs, and across both active and silent human epigenomic domains. Our analyses suggest that chromatin is comprised of a diverse array of both regular and irregular single-molecular oligonucleosome patterns that differ subtly in their relative abundance across epigenomic domains. This irregularity is particularly striking in constitutive heterochromatin, which has typically been viewed as a conformationally static entity. Our proof-of-concept study provides a powerful new methodology for studying nucleosome organization at a previously intractable resolution, and offers up new avenues for modeling and visualizing higher-order chromatin structure.