Ribosome quality control is a central protection mechanism for yeast exposed to deoxynivalenol and trichothecin

Network analysis of genetic interactions

To approach the basic processes encoded in such an assembly of genes with highly diverse functions we analysed their genetic interactions. Molecular functions are driven by the collaboration of genes usually represented in functional groups or pathways. To understand what these interactions are and how they are governed is a key question. One way of obtaining this information is to screen for genetic interactions, which underlie a large number of biological functions [37]. A network covering the genome-wide set of genetic interaction in yeast has been constructed [38], and allows for screening of interactions and clustered groups of interacting genes. Building upon this genetic interaction information we inferred four networks representing either a response to TTC, to DON, to either of these two (union) or both of them (intersection). To this end we used the corresponding set of genes as seeds and inferred the corresponding network from the list of genetic interactions. The resulting networks are provided as cytoscape file (Additional file 5: Data S4). At first we inspected the global clustering coefficient, which indicates the presence of local clusters within the networks that might encode for specific functionality. We found that yeast genetic interaction networks of DON or TTC responsive genes were more clustered than random networks (Additional file 6: Figure S1). This was also observed for the network of TTC or DON responsive genes. The shortest path between a pair of nodes is the minimal number of edges between them. Shorter path length might render networks more efficient at processing information than networks with another topology. This means essentially that a common connecting process exists in such a network. It has been suggested that small shortest path lengths might facilitate efficient communication between transcription factor genes [39]. The three networks (DON, TTC, DON or TTC) showed a significantly shorter average path length then their random counterparts (Additional file 7: Figure S2). The large global clustering coefficient and the short path length present typical properties of small world networks [40]. For the network diameter no such clear pattern was observed. An alternative approach for quantifying the network complexity based on topological properties is to make use of entropy-based network descriptors [41, 42]. We made use of a parametric graph entropy measure (Dehmer-entropy) which assigns a probability value to each node of a graph based on information functionals distributions [43]. These information functionals capture the structural information of a graph by considering a given probability distributions [43]. For DON, TTC, DON or TTC networks the distance measure of the sphere-based information functional distributions [43] was found to be significantly different from random networks (Additional file 8: Figure S3). This observation supports the idea that the toxin-responsive networks exhibit notable properties which render them efficient at processing information and thus collectively mediate resistance. In contrast to the topology of the DON, TTC, and DON or TTC (union) networks, the network of DON and TTC (intersection) responsive genes was not significantly different from random networks with the same number of seed genes. Genes representing the intersection of DON and TTC responsive genes might just represent indirect and pleiotropic effects of the respective mutants. One of the main functional clusters present in the intersection network (Fig. 3) was the Rpd3 histone deacetylase complex (HDAC), involved in transcriptional regulation.

Chromatin remodelling complexes are involved in DON/TTC resistance

The network analysis points out a HDAC and a few other highly contributing genes but also shows a tight interaction over all isolated genes. Thus a common theme is likely embedded in the ensemble of the isolated genes. In the following we argue that most mutants impair directly or indirectly the regulation of cytoplasmic protein synthesis or the production and maintenance of functional ribosomes. We concentrate the discussion on groups of mutants that are highly sensitive and the respective gene products are involved in the same pathway or in the same protein complex. In the group comprising mutants in chromatin metabolism and transcription, we identified several subunits (SDS3, SIN3, UME1, CTI6, TOD6, SAP30, RXT2) of the RPD3L Histone deacetylase complex and RCO1 of the RPD3S complex. The histone deacetylase Rpd3 is involved in repression of RNA Pol II driven rRNA synthesis in stationary phase [44, 45]. Moreover, loss of the Rpd3 histone deacetylase, in rpd3 mutants or in mutants lacking SIN3 or SAP30, results in a delay in rRNA processing but not in an rRNA transcriptional defect. Genetic evidence also points to rRNA posttranscriptional modification as a bottleneck in ribosome biogenesis [46]. This suggests that in Rpd3 mutants, de novo production of ribosomes might be reduced, a defect which may become more pronounced in DON and TTC treated cells due to additional inhibition of ribosomes. Alternatively, the lack of suppression of intragenic spurious transcription by the RCO1 mutant (RPD3S complex) might generate a burden by overloading RNA surveillance mechanisms [47]. Furthermore, we identified three subunits of the SWI/SNF chromatin remodelling complex (SNF2, SWI3, SNF6). SWI/SNF has been reported to play a role in transcription by RNA pol I in yeast. A 2.5-fold reduction in rRNA synthesis rate was reported for snf6Δ cells [48]. Moreover, SWI/SNF is required for repression of RNA polymerase II-dependent transcription in the ribosomal DNA (rDNA silencing) and telomeric silencing [49]. Thus reduced or aberrantly regulated rRNA synthesis in these and other loci (RRP8, REI1, PAF1, PIH1) might be one underlying reason for the increased DON sensitivity. Furthermore, we detected DON sensitivity in mutants lacking the histone deacetylases HDA1 or HDA3 (which interact within a complex) and the alternative H2A histone H2A.Z encoded by HTZ1. Their influence is unclear apart from the fact that basal transcription of ribosomal genes is repressed in the corresponding mutants. Another mutant in a gene with chromatin related function was the HMG protein Nhp6A which is recruiting the FACT complex (Spt16-Pob3) and other chromatin remodelling complexes. NHP6A is required for transcriptional initiation of tRNA genes and assembly of the RNA polymerase II and III pre-initiation complexes. Notably, also the CCR4-Not complex has been reported to be directly involved in rRNA synthesis [50].

Mitochondria are probably not a direct target of DON and TTC

Mitochondria have been implicated in the toxic effect of TTC [27, 28]. Genes connected to mitochondrial functions identified in this study do not seem to highlight a specific metabolic activity of mitochondria or mitochondrial translation. The relatively few mutants related to mitochondrial function (YMC1, DIC1, MBA1, UTH1, TOM5) and their low sensitivity does not support a model with mitochondria as major targets for both DON and TTC. The most sensitive mutant was lacking MDM32 a mitochondrial inner membrane protein required for normal mitochondrial morphology and inheritance. MDM32 is upstream of PNO1 an essential gene required for 18S rRNA processing. Interference of the knock out cassette with PNO1 expression under screening conditions cannot be excluded. The possibility remains that Pdr5-dependent drug efflux clears the cytosol of DON/TTC but not the mitochondrial matrix and is thus pronouncing effects connected to mitochondria as observed by other studies [28].

Cell integrity and cytoskeleton are indirectly involved in DON/TTC resistance

A substantial number of DON/TTC sensitive mutants were involved in cell integrity and cell wall biosynthesis. We identified several members of the protein kinase C (PKC) signalling pathway. These are BCK1 (MAPKKK), SLT2 (the MAP kinase), ZEO1 a regulator in the peripheral plasma membrane, and SLG1 a sensor-transducer. These mutants are multidrug sensitive [22, 36] and appear to be involved in many diverse processes such as regulating maintenance of cell wall integrity, cell cycle, mitophagy and pexophagy and organization of the actin cytoskeleton. For example, Slt2 plays a role in vacuole homeostasis and actin dynamics [51]. In addition to the multifaceted pathology of these mutants, the cell integrity pathway is also involved in repression of ribosome and tRNA synthesis in secretion-defective cells [52]. In addition, we found mutants affecting cell wall integrity or functionality like 1,3-beta-D-glucan synthase Fks2 or the alpha 1–6 Mannosyltransferases ANP1 and HOC1. Anp1 has been identified as conferring stress resistance in several instances [53, 54]. Cell wall synthesis is necessarily coupled to growth which in turn is dependent on biosynthesis. Alternatively, the yeast cell wall could have a protective role through drug adsorption properties. Yeast cell walls are sold as feed adjuvant with trichothecene protective activity [55]. β-glucan, is used to adsorbe DON [56, 57] and low but significant adsorption properties have been reported [56, 58]. Mutants impairing the function of the cytoskeleton are multi drug hypersensitive [36] and were found here to be highly sensitive to both DON and TTC. Their relative hypersensitivity score (MDR) [36] is listed accordingly (e.g. Table 1). We identified the interacting Actin-associated proteins Rvs167 and Rvs161 which have roles in endo- and exocytosis. Other sensitive mutants include Fimbrin (SAC6), Verprolin (VRP1) a proline-rich actin-associated protein involved in cytoskeletal organization and cytokinesis, and SLA1 required for assembly of the cortical actin cytoskeleton. Pbs2, the MAPKK of the high osmolarity glycerol (HOG) pathway has also a function in restructuration of the cytoskeleton [59]. A possible reason why such mutants appear in many screens is the observation that rRNA levels are repressed in various mutants with defects in transport and exocytosis [60]. The unfolded protein pathway is apparently not prominently involved in TTC resistance. The ire1 mutant was weakly sensitive to TTC and not to DON. Protein phosphatase 1 (PP1) is an enzyme clearly important for resistance against DON since we identified three genes belonging to PP1 interactors (REG1, GIP4, GAC1). The catalytic subunit of PP1 (GLC7) is an essential gene and therefore not included in the collection and functions to promote cell integrity, mitosis, bud morphology and polarization of the actin cytoskeleton. Thus PP1 might indirectly have a profound influence on ribosome synthesis.

Many mutants impair ribosome and ribosome quality control constituents

Genes functionally related to translation comprised several categories: ribosomal subunits (RPL39, RPL21A, RPS27B, RPS21A, RPS23B,), ribosome assembly (BUD20, CMS1, RPP1A, REI1, RRP8 rRNA methyltransferase), the Cop9 signalosome components (YJR084W, PCI8), initiation (SGN1, GCN3, GIS2, BUD27) and RNA turnover. Furthermore, we identified genes involved in RNA degradation such as LSM1, the Ski complex (SKI2, SKI8), and SLH1 (RNA helicase related to Ski2), and mRNA decapping (EDC3, PAT1), and splicing (BRR1, LEA2). Interstingly, bud morphology defects can partially be attributed to impaired ribosome function. A screen for abnormal bud morphology [61] detected 111 mutants including 15 genes (significance p = 0.0006) here identified as DON and/or TTC sensitive. These include functions in translation (RPS27B, RPL39, POP2, CCR4, BUD27, BUD20), cell wall (SLG1, PER1, FKS1, CCW12) and cytoskeleton (SLA1, RVS167, RVS161). Furthermore, DON and/or TTC sensitive mutants with defects in tRNA synthesis and processing involved uridine thiolation of cytoplasmic tRNAs (NCS6, URM1, IKI3), a non-canonical poly(A) polymerase which catalyses polyadenylation of hypo modified tRNAs, snoRNA and rRNA precursors (TRF5), and a Zinc-finger protein Stp3 possibly involved in pre-tRNA splicing. An interesting connection to possible oxidative stress caused by trichothecenes is the thioredoxin peroxidase Tsa1 which acts as ribosome-associated antioxidant [62]. However, other genes with dedicated antioxidant functions were not found. Biotransformation of trichothecenes occurs partially through glutathione addition [17]. A prominent influence on DON/TTC detoxification by formation of glutathione conjugates in yeast is not supported by our data. An exception is the prion encoded by URE2, which can mutate to a glutathione transferase but it is involved in growth related processes as well and thus the effect of the mutant might be rather indirect.

The mutant strain with the most striking sensitivity against DON was lacking HEL2/YDR266C. Hel2 is an E3 ligase and has been linked genetically with the E2 Ubc4. Hel2 has initially been implicated in histone turnover [63] and connected to translation and ribosome quality control (RQC) [64, 65]. Genetic data identified HEL2 as a high-copy-number suppressor of an allele of Mpe1, a conserved subunit of the cleavage and polyadenylation factor (CPF) [66]. Hel2 and Asc1 are factors for polypeptide quality control upstream of RQC [67, 68]. This is in line with the action of DON and TTC on early stages of translation. The RQC complex (Rqc1-Rkr1-Tae2-Cdc48-Npl4-Ufd1) is a ribosome-bound complex required for the degradation of polypeptides arising from stalled translation [67, 69]. We tested mutants lacking RQC1 and RKR1 and found no indication for increased DON or TTC sensitivity. Paromomycin is an aminoglycoside inducing the ribosome to bypass premature termination codons and to relax the stringency of decoding [70]. To investigate if DON/TCC toxicity is suppressed in the presence of paromomycin we determined the maximal growth rate in presence or absence of DON (12 mg/l) and a range of paromomycin concentrations (Additional file 9: Figure S4). We find a more than additive growth inhibition of the drug combination on pdr5Δ cells. Interestingly, we noted a higher resistance of the hel2Δ strain to higher doses (0.25 to 1 g/l) of paromomycin (Additional file 9: Figure S4A). Thus DON-induced stalling of ribosomes is not efficiently released by paromomycin-induced relaxation.

We confirmed the phenotype of the HEL2 deletion mutant by re-testing the original deletion strain of the ordered strain collection (Fig. 4a) and by deleting the HEL2 gene in a different strain background. For confirmation, we used a strain not related to the library strain and because of the low toxicity of DON lacking ABC transporters (Pdr5, Pdr10, Pdr15) as well as the yeast acetyl transferase Ayt1 [18]. The serial dilution tests showed a clear sensitivity of the respective HEL2 deletion mutants to DON and TTC. Since resolving of stalled ribosomes is an activity occurring also in unperturbed cells we expected that the HEL2 gene overexpressed under the control of the ADH1 promoter might improve the ribosome clearing capacity. Indeed, overexpression of Hel2 conferred a slight growth advantage in presence of DON in the growth test (Fig. 4b). Growth assays using different DON concentrations (Fig. 4c) showed a marked dose-dependent decrease of growth of the mutant but also a gain of resistance by overexpression of HEL2. This gain is detectable in the wild type and the complemented hel2Δ strain (Fig. 4c). Thus Hel2 becomes limiting in DON treated cells. Also the maximal growth rate was found to be reduced in the DON treated strain relative to the wild type (Fig. 4d).

Asc1, a homolog of human RACK1 (human receptor for activated C-kinase 1) and Hel2 have been implicated in translation through polybasic peptide sequences, and may act after the ribosome has translated some distance [67, 71]. Asc1, a core component of the small (40S) ribosomal subunit which prevents frameshifting of stalled ribosomes at repeats of CGA codons is related in its effective function to Hel2 [71, 72]. Translation of CGA codon repeats is inefficient, resulting in dose-dependent stalling of ribosomes [73]. In this respect it is interesting to note the TTC sensitivity of some mutants defective in arginine biosynthesis (ARG4, ARG5,6) as well as membrane transporters for cationic amino acids (RTC2, YPQ1, VBA5). These mutants could decrease the arginine pools and thus lower the availability of the rare tRNA-Arg(UCG). Also several methionine biosynthesis genes were identified (HOM6, MET12,13,14,22,31). The ASC1 gene contains an intron harbouring a snoRNA transcript unit (snoRNA U24). To distinguish between Asc1 and snoRNA dependent phenotypes we tested a mutant expressing the snoRNA U24 but not Asc1p due to a truncation of the first exon [74, 75] and found no effect from the U24 RNA (Fig. 4e). Thus the strong DON and TTC sensitivity of the mutant is due to the absence of Asc1. Asc1 is also required for general cell wall integrity and mutants are hypersensitive to iron starvation [74]. In that respect it is interesting to note that we detected genes involved in Iron metabolism (FRA1, FET4, NFU1, MTM1), as well as, a high-affinity copper transporter of the plasma membrane (CTR1) and ARR1 a transcriptional activator involved in resistance to arsenic compounds. The Rsp5 E3-ubiquitin ligase is involved in endocytosis, membrane protein trafficking and protein quality control and thus indirectly in many cellular processes but notably also in amino acid permease sorting [76, 77]. RSP5 is an essential gene and is therefore not present in the mutant collection but genes of two interacting proteins were identified. BUL2 coding for a component of the Rsp5p E3-ubiquitin ligase complex and RUP1 coding for a protein that regulates ubiquitination of Rsp5p. Taken together, we suspect that many of the isolated mutants ultimately influence cytoplasmic translation and thus cause synthetic phenotypes in the presence of DON and TTC.

Transcription profiling reveals demand for ribosomes and histones in DON treated cells

Gene expression pattern analysis is a powerful method to detect transcriptional responses and to deduce active signalling pathways in a given situation. Here we investigated whether specific intracellular responses are elicited by DON. We used 12 mg/l and 24 mg/l of DON which caused a slight growth delay and treated exponentially growing cells lacking Pdr5. DON was added when the cultures reached OD600 of 0.5 and cells were harvested when the cultures reached an OD600 of 1 or 1.5 and compared to untreated control cultures at the same optical density. To define the difference between short term and long term responses, we incubated cells with 24 mg/l DON overnight and re-grew them in fresh medium containing 24 mg/l DON to early exponential phase (OD600 of 0.1 to 1). All measurements were done in triplicate. Expression data were filtered and 382 genes were selected according to fold change values of above 0.5 or below−0.5 (log2) in at least one condition and analysed for GO term enrichment. We found enrichment of genes involved in mitochondrial respiratory chain function and amino acid biosynthesis, however, the distribution is almost even between the induced and repressed genes (Fig. 5a). Notably, the selected genes do not represent a specific cellular response. The gene lists and GO analysis are provided as supplementary data (Additional file 10: Data S5, Additional file 11: Data S6). We further analysed the absolute changes of signal abundance to define those transcripts which undergo the largest quantitative changes. The obtained expression level differences were normalized and arranged by hierarchical clustering (Fig. 5b). Significantly enriched GO terms connected to expression level differences revealed reduction of genes of glycolysis and amino acid biosynthesis and increase of genes involved in cytoplasmic translation. Interestingly, in this selection the levels of most ribosomal protein genes and all histone genes increased. Furthermore, expression level differences increase for most ribosomal protein genes (Fig. 5c, Additional file 12: Data S7). This possibly points to a situation caused by DON where translation is lacking behind the available carbon and energy sources. Reduced biomass production demand due to restricted growth leads to reduced glycolytic flux and a switch to respiration as energy source.

Many deletion strains sensitive to DON and TTC are sensitive to many other drugs as well [36]. Notorious multidrug sensitive strains have defects in functions of vacuole, cytoskeleton and cell wall. These mutants lead to a reduction of rRNA synthesis since growth and ribosome synthesis is tightly coupled [60]. One common response to environmental perturbations is the delay of the cell cycle progression. For example, various environmental stress types trigger depolarization of the actin cytoskeleton and cause a delay of the G2 phase of the cell cycle. On the other hand, mutations or drugs that specifically impair actin organization also trigger a G2 delay [78]. Many mutants with defects in various intracellular processes lead to a delay of G2. Both delay of the cell cycle and environmental stress lead to a characteristic gene expression pattern designated "slow growth signature" [79]. This transcript pattern is characterized by reduced expression of genes involved in protein synthesis such as ribosomal protein genes and glycolysis and the increase of transcripts of genes connected to respiratory growth. We collected the available microarray data for our DON and TTC sensitive mutants [79] and aligned the normalized DON treatment expression data (Figs. 6a and Additional file 13: Figure S5). The gene lists and values generating the heat maps are provided as supplementary data (Additional file 14: Data S8, Additional file 15: Data S9). We note a high proportion of strains exhibiting the slow growth signature (Pattern A) amongst our sensitive mutants suggesting that many mutant strains are suppressing ribosome production without treatment. Exposure to DON possibly leads to further reduced availability of translation capacity in these mutants due to degradation and stalling of translating ribosomes. CTI6, PDE2, YKR023W, VAC8, URM1, IRE1, APQ12, NTH2, URE2, VHR2, and RTG1 have the inverse transcript pattern (Pattern B). The comparison of the transcript profiles of DON-treated cells to the compendium pattern shows a slow growth signature for the early time points and a switch to the inverse pattern at later time points and during chronic exposure (Fig. 6b). Taken together this implies a reduced ribosome synthesis rate in many mutants which might be exacerbated by DON induced stalling during translation.