Methionine restriction and cancer treatment: a systems biology study of yeast to investigate the possible key players

Background/aim Dietary restriction, mainly carbon and/or methionine restriction are among the upcoming supporting interventions along with chemotherapy in various cancers. Although dietary restriction has been proven to be beneficial, the main cellular machineries affected by its administration lacks deeper information considerably, a notable pitfall in its use as a personalized nutritional approach. Materials and methods In this study, cellular effects of methionine restriction on a yeast model are explored via systems biology approaches. The methionine biosynthesis network, constructed by integrating interaction data with gene ontology terms, was analysed topologically, and proved to be informative about the intertwined relationship of methionine biosynthesis and cancer. Experimentally, effects of methionine restriction on the yeast model were explored in vivo, with transcriptome analyses. Results The integrative analysis of the transcriptional data together with the reconstructed network gave insight into cellular machineries such as TOR, MAPK, and sphingolipid-mediated signaling cascades as the mostly responsive cellular pathways in the methionine-restricted cases with Sch9p (functional orthologue of mammalian S6 kinase) being placed at the intersection of these signaling routes.

(polyamine synthesis), and effects on DNA methylation (S-adenosyl methionine synthesis) (Wanders et al., 2020). Recent studies demonstrating the antiproliferative role of MR on colorectal , triple negative breast (Jeon et al., 2016), and prostate (Lu et al., 2020) cancer cells further fortify the use of MR as a potential therapeutic agent.
Saccharomyces cerevisiae, the budding yeast, has long been used as a model organism for human cancers, both for the broader investigation of cellular machineries leading to cancer biogenesis (Cazzanelli et al., 2018) and for studying the response to anticancer agents (Matuo et al., 2012). Effects of MR have also been studied in yeast, although research was limited mainly to the field of yeast aging (Ruckenstuhl et al., 2014;Plummer and Johnson, 2019). The transcriptional and translational dynamics of Saccharomyces cerevisiae cells upon MR have also been explored via RNA-Seq and ribosome profiling recently (Zou et al., 2017), focusing mainly on the translational regulation by MR.
The current study is the first study where the integration of transcriptome and interactome data for yeast, by systems biology approaches, is conducted. The methionine biosynthesis network reconstructed in S. cerevisiae is used as a backbone to investigate potential alterations and crosstalks with other signaling networks when methionine is limited. The integrative analysis provides key proteins, processes, and/or pathways having a role in the demonstrated beneficial effect of MR on cancer treatment.

Reconstruction of the methionine biosynthesis network
The methionine biosynthesis network was reconstructed via the Selective Permissibility Algorithm (SPA) (Arga et al., 2007;Börklü-Yücel and Ülgen, 2011). The 8 (HOM2p,HOM3p,MET6p,MET13p,STR3p,HOM6p,MET2p,MET22p) proteins which were manually curated to have the "methionine biosynthetic process" in their GO function term on Saccharomyces Genome Database (SGD) as of January 2023, were adopted as the seed proteins from which the network expands. The Annotation Collection table, the second input for the reconstruction algorithm, was created by pooling the process, function, and component GO annotations of the mentioned seed proteins only, downloaded from SGD, as of January 2023. The third input for SPA is the physical protein-protein interaction file obtained from BIOGRID, release 4.4.218. Briefly, the reconstructed network expands from these 8 seed proteins in such a way that a candidate protein is selected to be a member of the network based on its GO Terms and its physical interaction data: it is included in the network if and only if its three GO Terms are present in the Annotation Table  and it interacts physically with one of the seed proteins. The included proteins constitute the "first neighbors" of the seed proteins, and the network continues to expand by taking these first neighbors as the new seed proteins for the rest of the automated algorithm until no new proteins are found to be added to the network.

Network analysis
For the topological analysis of the reconstructed network, the "NetworkAnalyzer" tool of Cytoscape software was used (Assenov et al., 2008). Highly densely connected proteins of the reconstructed network were identified with MCODE (Bader and Hogue, 2003) algorithm. Responsive subnetworks were created with "jactivemodules" plugin of Cytoscape, with the default settings (Ideker et al., 2002). Briefly, a z-score for a subnetwork A which has k members is calculated according to z(A) =

Network analysis
For the topological analysis of the reconstructed network, the "NetworkAnalyzer" tool of Cytoscape were created with "jactivemodules" plugin of Cytoscape where ysis lysis of the reconstructed network, the "NetworkAnalyzer" tool of Cytoscape ivemodules" plugin of Cytoscape ( ) = −1 (1 − ) φler yapılacaktı.
1 √ ∑ ∈ = −1 (1 − ) , with φ -1 being the inverse of the normal cumulative distribution function and p being the p-values of the nodes present in the subnetwork. The module score s(A) for the subnetwork A is the normalized value of this z-score with mean and standard deviation of scores z (A) over all k-node groups of the graph.

Experimental procedure
The strain used in the experiments was ΔHO derived from BY4742 background, Matα; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0; YDL227c::kanMX4 obtained from EUROSCARF deletion collection. For the main cultures, overnight cultures grown in SDC media were diluted to an OD600 value of 0.1 and inoculated into fresh SDC media in micro-aerated flasks, with a working volume of 1:5. The cultures were then grown batch-wise at 30 °C and 180 rpm up to midexponential phase (OD600 ~ 0.6). At the mid-exponential phase, cells of the main culture were divided into three aliquots which were centrifuged at 6000 rpm for 5 min, washed twice with deionized, and distilled sterile water prior to their transfer in the treatment media. Treatment media comprised of fresh SDC (2% methionine) for the control case, SDC + 0.75% methionine for the methionine restricted case. Samples for transcriptome analyses were collected 2 h after the treatment. The experiments were done in biological triplicates.

Sampling and mRNA extraction
For mRNA extraction, 5 mL samples were collected, frozen in liquid nitrogen immediately, and stored at -80 °C prior to RNA extraction. RNA extraction was performed automatically with QiaCube using RNeasy Mini Kit (Quiagen), as described by the manufacturer.

Microarray data acquisition and processing
The qualitative and quantitative spectrophotometric analysis of RNA was done using UV-vis spectrophotometer (NanoDrop ND-1000, Thermo Fisher Scientific Inc., U.S.A). RNA integrity number (RIN) values were checked using a microfluidics-based platform (Bioanalyzer 2100 Agilent Technologies, USA) using RNA6000 Nanokit (Agilent Technologies, USA) and samples with RIN values 7-10 were processed. The microarray analysis steps comprising of the synthesis of cDNA, conversion of cDNA into a double-stranded DNA, transcription and synthesis of biotin-labeled aRNA from the double stranded DNA, purification and fragmentation of aRNA, and the final hybridization of aRNA were performed as described in the Affymetrix GeneChip®Expression Analysis Technical Manual.
Data were processed in R using raw cell files via the "affy" (Gautier et al., 2004) package of Bioconductor, with "rma" option chosen for background correction and "quantile" option for quantile normalization. Final expression values for the genes were obtained in log2 transformed values, differentially expressed genes were obtained by Student's t-test GO Term enrichment results of the differentially expressed genes were obtained via web-based "gprofiler" server (Raudvere et al., 2019), with a Benjamini-Hochberg FDR value of 0.05 as the significance threshold. Data were submitted to ArrayExpress database, under the accession number E-MTAB-12860.

The key proteins of the methionine biosynthesis network are linked to cancer
The methionine biosynthesis network, reconstructed with SPA, had the 8 genes which had the "methionine biosynthetic process" as their manually curated GO process term as the seed genes (Table). The annotation table used to expand the network, pooled terms of these seed proteins, comprised of 7 component GO terms together with 31 function and 29 process terms (Table). The final undirected network reconstructed had 1102 nodes and 10141 edges (Figure 1a). The topological analysis of the network revealed that it reflected the "small world" property of the biological networks, with network diameter and mean path length values of 5 and 2.7 respectively, despite the large node number of 1102. Moreover, the distribution of node degrees follows nearly a power law model (P(k) = k -γ ) as in many other biological networks, with γ = 0.895 and an R 2 value of 0.77 ( Figure  1b). These parameters imply that the network in question contains numerous small, highly integrated modules as is the case for many other active biological networks.
The highly connected nodes, referred to as the "hubs" of the network, reflect the highly impactful genes in the network, around which various other system components (nodes) revolve. On the other hand, the nodes with high "betweenness centrality" (BC) values are the nodes through which shortest paths are traversed more heavily, i.e. they contribute more to the information flow in the network. With these descriptions in hand, the top 10 hubs and top 10 nodes with the highest BC values of the network are inspected closely (Figure 1c). A combination of these two lists resulted in 13 genes (CDC28, SGV1, PKC1, DBF2, SLT2, ELM1, THR4, PTK2, SNF1, ILV1, CDC8, NOP2, and TTI1) each with a human orthologue, known to be implied in various cancers ( Figure S1).
Cdc28p (yeast orthologue of human CDK1) is a cyclin-dependent kinase controlling the mitotic cell cycle. Apart from the tight relationship between cell cycle abnormalities and senescence of cancer cells, CDK1 upregulation is known to be related to reduced survival time for colorectal cancer, liver cancer, and lung cancer (Kuang et al., 2019). Sgv1p, CDK9 in Homo sapiens, is another cyclin-dependent kinase, involved in the regulation of transcription elongation from RNA polymerase II promoters. Recent studies demonstrated that the inhibition of CDK9's activity holds the potential to be a highly effective anticancer therapeutic (Mandal et al., 2021). Pkc1p, orthologue of human PKC, is a serinethreonine protein kinase proven to have important roles in carcinogenesis and can serve as a promising target for cancer treatment (He et al., 2022). STK38 is the human orthologue of another Ser/Thr kinase of the network, Dbf2p, which has a significant prognostic value in different cancers and is closely associated with cancer immunity . Slt2p is a member of the MAPK cascade in yeast, as its orthologue ERK5 in humans, and alterations of the MEK5/ERK5 signaling pathway have been described in several tumor cells (Monti et al., 2022). Members of another signaling pathway, mammalian AMPK signaling, are also among the key proteins: Snf1p and Elm1p. Snf1p (AMPK) along with its regulator Elm1p (CAMKK), is another promising potential target for cancer prevention and treatment (Li et al., 2015). Ptk2p, the yeast orthologue of human MARK2, is another protein kinase of the network. Overexpression of MARK2 in lung tumor cohorts is well established and novel roles in pancreatic cancers are also described for MARK2 (Zeng et al., 2022). Cdc8p, another key kinase, is the yeast orthologue of human DTYMK whose inhibition has been shown to restrain the growth of hepatocellular carcinoma and increase sensitivity to the therapeutic agent oxaliplatin . Nop2p is a methyltransferase whose human orthologue NOP2 has been demonstrated to be upregulated in colon cancer (Bi et al., 2019). Tti1p is a telomere length regulator, with the orthologue of TTI1 in humans. Overexpression of TTI1 in colon and lung cancer has recently been unraveled (Xu et al., 2021). The remaining two key proteins are directly related to threonine metabolism, Thr4p (THNSL2), and Ilv1p (SDS). Although a definite role for threonine metabolism in cancer has not yet been proposed, there are theories about fueling of the TCA cycle by threonine which results in the release of ATP for the required energy for oncogenic activities (Lieu et al., 2020).
The tight interconnection of these 13 hub proteins with various cancer types, has proven that the dynamics of the methionine biosynthetic process may indeed be one of the key factors in cancer biology.

Microarray analysis reveals that methionine restriction triggers sulfate assimilation while represses ribosome biogenesis pathways
Statistical analysis of the transcriptome data of methionine restriction vs. control case resulted in 247 differentially expressed genes (FDR p-value < 0.05, absolute fold change >1.5) of which 121 and 126 were down-and up-regulated in methionine restriction case, respectively (Figure 2, Table S1). The GO Term and KEGG pathway enrichment results of these differentially expressed genes gave two prominent results: Genes downregulated in methionine restriction case were primarily enriched with "Ribosome biogenesis" and "rRNA methylation" (p-val:2.54E-17) terms. This result means that in the dietary restricted case, ribosome synthesis and/or assembly was down-regulated probably due to the methylation defects of rRNA, since methionine is scarce. This fact may be one of the beneficial effects of methionine restriction on cancer cells since   uncontrollably proliferating cancer cells need to produce ribosomes to sustain continuous proliferation and expand in numbers. In fact, recent work stressed the importance of targeting ribosome biogenesis as a potential therapeutic intervention in cancer (Zisi et al., 2022). The upregulated genes were enriched in "sulfate assimilation" (p-val:1.64E-11) pathway. When this pathway is closely scrutinized, it is seen that all the genes of the pathway are significantly upregulated in methionine restriction (Figure 3). The end product of this pathway, however, hydrogen sulfide, is a gasotransmitter and has been reported to have contradictory effects related to cancer biogenesis and progression. This controversial role of H 2 S in the cancer research field is explained by a bellshaped (biphasic) model, in which lower concentrations of H 2 S display procancer effects while higher concentrations exhibit anticancer properties (Hellmich and Szabo, 2015). It is proposed in this study that methionine restriction increases endogenous H 2 S levels to a threshold level that exhibits anticancer properties. It is known that yeast cells grown in methionine restricted conditions prefer quiescence over senescence, and this fact occurs simultaneously with increased hydrogen sulfide production (Choi et al., 2019). To evade senescence with higher H 2 S production may support this level-altering hypothesis, although further experimentation is needed for a definite result.
Results of the microarray experiments are in line with the literature: according to Zou et al.'s study also, upregulated genes are enriched in the amino acid/ methionine biosynthesis pathways whereas downregulated genes are enriched with terms pertinent to ribosomal processes (Zou et al., 2017). MET1, MET3, MUP3, SUL1, and SUL2 are all among the commonly upregulated genes in both studies, genes involved in methionine biosynthesis. ATG41 and INO1 however, are among the significantly upregulated genes in this study only, hinting growth phase at which methionine is restricted may affect the transcriptional profiles.
Atg41p's function is unknown but is known to be induced in autophagy-inducing conditions (Yao et al., 2015) and is required for autophagy and mitophagy. This result may point to ATG41 for the lifespan-increasing role of MR via mitophagy (Plummer and Johnson, 2019). The tumor-suppressing role of mitophagy in the early stages of tumor development is also well-established (Ferro et al., 2020). Ino1p on the other hand, is involved in the synthesis of inositol phosphates and inositol-containing phospholipids, especially IP 6 , which in its turn enhances the anticancer effect of conventional chemotherapy, controls cancer metastases, and improves the quality of life in cancer patients (Vucenik et al., 2020). Thus, among other signaling pathways, another signaling pathway tightly intertwined with methionine restriction may be inositol-phosphate signaling machinery, according to the transcriptome analysis.

Integration of transcriptome data with methionine biosynthesis network points to Sch9p as a key regulator between MAPK, TOR, and sphingolipid signaling machineries in response to methionine restriction
To dynamically examine the reconstructed network in part 3.1, "jactivemodules" plugin of Cytoscape is adopted. Briefly, gene expression data of part 3.2 is integrated into the methionine biosynthesis network, yielding an active subnetwork which captures the dynamic information correlated with methionine restriction. Computational integration of the network and the transcriptomics profile leads to the extraction of context-dependent active modules, which mark regions of the network showing striking changes in molecular activity associated with methionine restriction.
The integration procedure resulted in a small subnetwork of 243 nodes and 1442 edges. This responsive subnetwork was then processed via MCODE algorithm to discover the densely connected components: only 3 modules with a score of >3 are found. These highly interconnected regions represent protein complexes and/ or parts of a pathway which are triggered in methionine restricted cases. The combined analysis of these three modules yielded a single connected component, a dense responsive network of 53 nodes and 393 edges, referred to henceforth as the "dense network". Enrichment results of this dense network gave insight into the cellular machineries affected in the methioninerestricted case. According to this analysis, methionine restriction affects TOR and MAPK signaling pathways along with sphingolipid biosynthetic process (Figures 4 and 5). A detailed, closer inspection of the members of this dense network provides a comprehensive interpretation of systems altered in methionine restriction.
Tor1p and Tor2p, orthologues of mammalian mTOR, are already expected to be one of the key nodes of the dense network since amino acid limitation has a direct impact on TOR signaling to regulate growth in response to nutrient stress in Saccharomyces cerevisiae. A similar phenomenon has also been described in Homo sapiens (Takahara et al., 2020). Fus3p, the yeast orthologue of human MAPK1 (ERK5), is another member of the dense network, and it was published that a hyperactive ERK5 has a direct role in tumorigenesis (Guo et al., 2020). Three enzymes of the network are playing a part in sphingolipid biosynthesis, thus sphingolipid-mediated signaling: Inp52p, Lcb2p, and Sch9p. Inp52p takes part in phosphatidylinositol (PI) production while Lcb2p catalyzes the formation of 3-dehydrosphinganine, the rate-limiting step in sphingolipid biosynthesis. Sphingolipids are bioactive molecules which have signaling properties and are recently been investigated for their potential roles in cancer and therapy . Sch9p, the functional orthologue of mammalian S6 kinase, situates itself at the junction of these pathways (Figure 4): S6 kinase is phosphorylated by mTOR and ERK while it is inhibited by the sphingolipid signaling molecule ceramide (Wang et al., 2001;Magnuson et al., 2012;Jesko et al., 2019). It may thus be hypothesized that Sch9p integrates the information from TOR, MAPK, and sphingolipid-mediated signaling pathways in response to methionine restriction and subsequent beneficial antitumor effects.

Conclusion
Dietary restrictions emerged as new nutritional interventions which may enhance the efficacy of chemotherapeutic agents. Results of the current study revealed that the reconstructed methionine biosynthesis network has cancer-related orthologues as the hub nodes, fortifying the potential use of methionine restriction as a subsidiary intervention to fight cancer. Microarray analysis of cells subjected to methionine restriction pointed to hydrogen sulfide production as the triggered cellular machinery while ribosome synthesis was repressed. Both phenomena have been mentioned as therapeutic targets for various cancer types. Integration of transcriptome data with the reconstructed network broadened the perception of various signaling machineries such as TOR, MAPK, and sphingolipid-mediated pathways, each triggered in response to methionine restriction. Sch9p, the yeast orthologue of human S6K, seems to be the key player in the mentioned beneficial cellular response by taking place in the crosstalk between these different signaling routes. Further experiments which solely target S6K phosphorylation, or target S6K in combination with dietary restriction will provide valuable insights for antitumor therapy.

Supplementary data
Supplementary data can be accessed at the following link: https://aperta.ulakbim.gov.tr/record/252436 Acknowledgment/disclaimers/conflict of interest I thank KUTTAM and Koç University School of Medicine for their continuous support. I declare no conflict of interest.

Conflict of interest
The author has no conflicts of interest to declare. Figure S1. The visualization of 13 hub genes and their first interacting partners in the network. Hub genes are in blue while the interaction partners are in pink.