A yeast chemogenomic screen identifies pathways that modulate adipic acid toxicity

Summary Adipic acid production by yeast fermentation is gaining attention as a renewable source of platform chemicals for making nylon products. However, adipic acid toxicity inhibits yeast growth and fermentation. Here, we performed a chemogenomic screen in Saccharomyces cerevisiae to understand the cellular basis of adipic acid toxicity. Our screen revealed that KGD1 (a key gene in the tricarboxylic acid cycle) deletion improved tolerance to adipic acid and its toxic precursor, catechol. Conversely, disrupting ergosterol biosynthesis as well as protein trafficking and vacuolar transport resulted in adipic acid hypersensitivity. Notably, we show that adipic acid disrupts the Membrane Compartment of Can1 (MCC) on the plasma membrane and impacts endocytosis. This was evidenced by the rapid internalization of Can1 for vacuolar degradation. As ergosterol is an essential component of the MCC and protein trafficking mechanisms are required for endocytosis, we highlight the importance of these cellular processes in modulating adipic acid toxicity.


INTRODUCTION
Adipic acid has been described as the most important dicarboxylic acid as it is a valuable platform chemical for the production of nylon-6,6 (Polen et al., 2013). The global need for adipic acid has been projected to increase because it can be converted to other products including plasticizers, plastics, lubricants, and a food ingredient in gelatin (Musser, 2000;Polen et al., 2013;Weber et al., 2012). Nearly all of the world's adipic acid is currently produced from a chemical process that involves the conversion of benzene to adipic acid through several chemical reactions requiring an oxidation reaction of nitric acid and air (Musser, 2000). These chemicals are costly and toxic to the environment and the nitric acid oxidation step results in the production of nitrous oxide gas, a greenhouse gas contributing to global warming and thinning of the ozone layer. Indeed, it is estimated that about 10% of the worldwide nitrous oxide gas released yearly into the atmosphere comes from adipic acid production (Alini et al., 2007).
Considering these challenges in the chemical synthesis of adipic acid, there has been a huge drive toward a more sustainable and environment friendly production of adipic acid by microbial fermentation. In this bioprocess, microorganisms such as bacteria and yeast are engineered to convert glucose to adipic acid. Yeasts, particularly Saccharomyces cerevisiae, are a preferred microbial host for adipic acid fermentation via the shikimate pathway, primarily because yeast metabolism has been extensively characterized and is easy to engineer (Averesch and Krö mer, 2018;Nevoigt, 2008). Also, yeasts have previously been shown to be more tolerant to adipic acid toxicity than bacteria (Karlsson et al., 2017) and S. cerevisiae is generally robust to industrial stresses including low pH conditions (Chen et al., 2009;Fletcher et al., 2017), making it suitable for the production of organic acids including adipic acid. Although progress has steadily been made to produce adipic acid using engineered yeasts (Raj et al., 2018), the process is constrained by the toxicity of adipic acid to yeast. Hence, product toxicity is a key factor that should be considered during the design of engineered strains as this will greatly impact the final titers and productivity of the strain during the industrial production of adipic acid on a commercial scale.
Currently, the exact mechanism of adipic acid toxicity is unclear but is likely similar to the generic mechanism for organic acid toxicity. At a pH lower than the pKa of the acid (adipic acid pKa1 = 4.41, pKa2 = 5.41), organic acids freely diffuse into the cell where they then dissociate in the cytosol at a pH $7.0 into the anionic form, which is unable to diffuse back out of the cell due to the hydrophilic nature of the anions (Jarboe et al., 2013). In order to prevent accumulation and toxicity of the anionic form of the acid, the cell Chemogenomic screens have previously been used to identify mechanisms of toxicity not only for organic acids (Hazelwood et al., 2010;Kawahata et al., 2006;Mira et al., 2009;Piotrowski et al., 2015; but also for several other toxic compounds including fermentation inhibitors Skerker et al., 2013). This screen allows the identification of hypersensitive mutants that show significant growth defects in the presence of a toxic compound. It also allows the identification of hypertolerant mutants lacking a suppressor gene that when deleted results in improved growth in the presence of the inhibitor compound. Despite the promising potential of adipic acid production by yeast fermentation, no previous study has reported specific mechanisms underlying the acid's toxicity to yeast. Here, we utilized a yeast chemogenomic screen to identify unique biological processes that modulate adipic acid toxicity, which can be potentially targeted to engineer tolerance in yeast.

Chemogenomic screen reveals a role for ergosterol and protein trafficking in adipic acid tolerance
In order to identify biological processes underlying adipic acid toxicity in yeast, we performed a chemogenomic screen using the S. cerevisiae haploid deletion mutant array (DMA) collection to identify mutants that are hypersensitive or have increased tolerance to a sublethal concentration (80 mM) of adipic acid. The yeast DMA collection was pinned onto YPD agar plates containing 80 mM adipic acid (dissolved in DMSO), which lowered the pH of the agar plates to 4.0. To specifically identify mutants with modulated growth on adipic acid and not solely decreased pH, as a control the DMA collection was also pinned onto YPD plates containing DMSO (solvent control) and buffered at pH 4.0. We incubated the plates for 48 h after which the colony sizes of mutants grown on the adipic acid plates were compared with those of the pH 4.0 control plates and were analyzed using the synthetic genetic array (SGA) tools . We identified a total of 77 adipic acid-sensitive mutants (SGA cutoff < À0.3) and one mutant that conferred improved growth on adipic acid (Table S1). To eliminate any false-positives, we quantified the growth rate of the 77 mutants in YPD medium supplemented with adipic at pH 4.0. The ratio of growth rate in the presence of adipic acid to that in the absence of adipic acid was obtained as the PRECOG score ll OPEN ACCESS 2 iScience 24, 102327, April 23, 2021 iScience Article , which represented the level of sensitivity or tolerance of the mutants to adipic acid. Using a PRECOG score of R2, signifying more than 50% growth inhibition in the presence of adipic acid, we confirmed 43 mutants to be sensitive to adipic acid. In contrast only one suppressor or deletion mutant was confirmed to grow better in adipic acid (PRECOG score %0.8).
To further validate that our screen specifically selected mutants sensitive to adipic acid and not to the low pH (pH 4.0) conferred by adipic acid, we compared our data with a previously reported chemogenomic screen for pH 4.0 toxicity (Shin et al., 2016). We found no overlap between our data and the 129 genes identified in the low pH screen (Shin et al., 2016) (Table S1), giving a strong indication that the sensitive strains obtained in our screen have been selected for specific sensitivity to adipic acid. Previous work had reported that overexpression of PDR12 improved tolerance to adipic acid when overexpressed (Pereira et al., 2019), and our screen identified pdr12D cells as being hypersensitive to adipic acid, providing further evidence that the chemogenomic screen is specific to adipic acid.
Gene Ontology (GO) enrichment analysis  identified numerous biological processes that were enriched in the screen including the ergosterol biosynthesis pathway (p = 1.147 3 10 À5 (Table S1; Figure 1). Indeed, strains with an ERG2, ERG3, or ERG4 deletion were hypersensitive to adipic acid and showed no growth in the presence of 120 mM adipic acid, suggesting that ergosterol biosynthesis is required for growth in adipic acid ( Figure S1). The most represented biological process that modulates The colored nodes depict genes that result in adipic acid sensitivity in yeast when deleted. The different genes are clustered based on color-coded biological processes. Adipic acid sensitivity resulting from gene deletion based on calculated PRECOG scores is shown by the node sizes. The suppressor gene is indicated by a star. Gene nodes connected by a red line represent genes encoding components of a protein complex as indicated. See also Figure S1.

OPEN ACCESS
iScience 24, 102327, April 23, 2021 3 iScience Article adipic acid toxicity in our screen is protein transport and trafficking (p = 3.80 3 10 À11 ; Table S1 and Figure 1). In particular, genes encoding proteins within the ESCRT pathway such as STP22-VPS29 (ESCRT-I), VPS25-SNF8-VPS36 (ESCRT-II), and SNF7 (ESCRT-III) were essential for growth in the presence of 120 mM adipic acid. Indeed, all these ESCRT pathway mutants failed to grow in the presence of 120 mM adipic acid. Deletion of Bro1, an auxiliary protein in the ESCRT pathway, which binds to SNF7 (Wemmer et al., 2011), also results in sensitivity to adipic acid. Furthermore, deleting genes encoding protein complexes such as VPS51-VPS53 (GARP complex) and COG5-COG6-COG7 (COG complex) results in significant growth inhibition in the presence of adipic acid (Figure 1). Taken together, ergosterol biosynthesis and protein trafficking are essential biological processes required for yeast growth in adipic acid as deletion of key genes in these processes results in hypersensitivity to adipic acid.

KGD1 in the tricarboxylic acid cycle suppresses adipic acid toxicity
The only adipic acid suppressor mutant identified was KGD1, which encodes an enzyme (a-ketoglutarate dehydrogenase) that converts a-ketoglutarate to oxaloacetate (Repetto and Tzagoloff, 1989) in the tricarboxylic acid (TCA) cycle ( Figure 2A). We confirmed that the kgd1D strain grew better than the wild type during exposure to adipic acid in both liquid and solid YPD media (Figures 2B and S1A). Through liquid growth assays, we next investigated the involvement of other genes in the TCA cycle in modulating adipic acid tolerance. Apart from sdh1D, which is hypersensitive to adipic acid, all the other TCA cycle mutants grew similarly as the wild-type strain in adipic acid ( Figure S2B).
Next, we asked if deleting KGD1 also improved tolerance to catechol and cis,cis-muconic acid, two precursors in the adipic acid biosynthetic pathway. Therefore, we calculated yeast growth rates in liquid cultures containing 12 mM catechol or 50 mM cis,cis-muconic acid and compared it with that grown in YPD without any chemicals added. Although the kgd1D strain grew better than the wild type in media containing catechol (p < 0.05), the mutant strain grew similarly as the wild-type strain in media containing cis,cis-muconic acid ( Figure 2C), suggesting that KGD1 deletion confers cross-resistance to some precursors in the adipic acid biosynthesis pathway. In summary, KGD1, and not the TCA cycle, modulates adipic acid tolerance in yeast. Furthermore, deletion of KGD1 also improved tolerance to catechol, a toxic adipic acid precursor.
Genes encoding enzymes in the last four steps of the ergosterol biosynthesis pathway are required for growth in adipic acid To determine whether the ergosterol biosynthesis pathway ( Figure 3A) modulates adipic acid toxicity, using liquid growth analysis we tested the sensitivity of all nine non-essential ergosterol pathway deletion mutants to adipic acid (HMG1, HMG2, ERG24, ERG6, ERG28, ERG2, ERG3, ERG5, ERG4) (Bhattacharya et al., 2018). These included the last four genes in the ergosterol biosynthesis pathway that were identified in our chemogenomic screen (ERG2, ERG3, ERG5, and ERG4, Figure 1). Similar to our chemogenomic screen, whereas erg5D cells display a mild sensitivity to adipic acid, erg2D, erg3D, and erg4D cells showed a complete growth inhibition in media containing 120 mM adipic acid ( Figure 3B). However, whereas hmg2D cells have minor sensitivity to adipic acid at higher OD, the other non-essential ergosterol pathway deletion mutants cells' growth profile was similar to those of wild-type cells upon 120 mM adipic acid treatment ( Figure 3B).
In order to delve deeper into the role of the ergosterol pathway in modulating adipic acid toxicity, including the essential pathway proteins, we next asked if 2 h of adipic acid treatment impacted the protein levels and/or subcellular localization of ergosterol pathway enzymes tagged with a green fluorescent protein (GFP). We did not observe any significant changes in protein level of any of the GFP-tagged ergosterol pathway enzymes upon adipic acid treatment ( Figures S3 and S4). This suggests that even though the ergosterol biosynthesis pathway, particularly the last four steps, is required for growth in adipic acid, the pathway is not induced in the presence of the compound. However, two of the enzymes in the pathway changed localization upon adipic acid treatment. The endoplasmic reticulum (ER) localized Erg24-GFP (Jordá and Puig, 2020) and Erg4-GFP (Srinivas et al., 2019) moved to the cytosol upon adipic acid treatment ( Figure 4A). Erg4-GFP also changed localization to punctate structures ( Figure 4A). As Erg4 is the last enzyme in the pathway leading to the formation of ergosterol and its deletion results in hypersensitivity to adipic acid, we did a time course experiment to track changes in localization of Erg4-GFP. To this end, we grew cells expressing Erg4-GFP to log phase in YPD before adding 120 mM adipic acid, and localization was assessed at 0, 15, 30, and 60 min. We found that Erg4-GFP changed localization to punctate structures as early as 15 min of adipic acid exposure ( Figure S5). As ergosterol is a major component of The images are representative of three biological replicates.
(C) Wild-type (WT) and kgd1D (YKB4943) strains were grown to the mid-log phase; diluted to an OD 600 of 0.1 in liquid YPD cultures with or without 12 mM catechol, 50 mM muconic acid, or 140 mM adipic acid; and automated growth curves were obtained in triplicates using the Bioscreen at 30 C. Growth rates of kgd1D were calculated for each treatment and normalized to that of the WT. Error bars represent 1 standard deviation. *p < 0.05. See also Figure S2.

OPEN ACCESS
iScience 24, 102327, April 23, 2021 5 iScience Article mitochondrial membranes (Krumpe et al., 2012) and decreases in pH can cause fragmentation of mitochondria (Pereira et al., 2010), using the mitochondrial marker Cit1-RFP, we asked if Erg4-GFP was relocalizing to mitochondria upon adipic acid treatment. As expected, in untreated cells, the ER-localized Erg4-GFP did not co-localize with Cit1-RFP ( Figure 4B). Adipic acid treatment caused fragmentation of the mitochondria, and Erg4-GFP co-localized with Cit1-RFP. Together, these data suggest that the last four steps in the ergosterol biosynthesis pathway that convert fecosterol to ergosterol, although not induced during adipic acid treatment, are required for yeast growth during exposure to adipic acid. Indeed, key enzymes such as Hmg1, which catalyzes a rate-limiting step of the pathway (Bhattacharya et al., 2018), and Erg4, which catalyzes the final step in the pathway, change localization upon adipic acid treatment, suggesting a potential regulation of the pathway during growth in adipic acid.
Ergosterol and Pdr12 efflux play non-overlapping roles protecting the cell from adipic acid toxicity Plasma membrane ergosterol has been implicated in the localization and activity of many plasma membrane transporters (Bagnat et al., 2000;Kodedová and Sychrová , 2015). Hence, we next sought to determine if the hypersensitivity of ergosterol biosynthetic pathway mutants to adipic acid reflected defects , and erg4D (YKB4876) cultures were grown to the mid-log phase, diluted to an OD 600 of 0.1 in liquid YPD cultures with or without 120 mM adipic acid, and automated growth curves were obtained in triplicates using the Bioscreen at 30 C. The growth curves shown represent an average of three replicates. See also Figure S1. iScience Article in Pdr12, the only plasma membrane transporter identified in our chemogenomic screen (Figures 1 and S6).
As the overexpression of PDR12 improves adipic acid tolerance (Pereira et al., 2019), we first asked if adipic acid treatment changed the protein levels or localization of Pdr12. We C terminally tagged endogenous PDR12 with GFP (Pdr12-GFP) and assessed the impact of adipic acid treatment on protein levels and localization. Upon 120 mM adipic acid exposure, there was a rapid induction of Pdr12-GFP, which localized to the plasma membrane ( Figure 5A). The induction of the protein was observed as early as 15 min after exposure to adipic acid ( Figure 5A). Using quantitative western blots we found that Pdr12-GFP protein (C) Deletion of ERG2 and ERG4 does not impact the adipic acid induction and plasma membrane localization of Pdr12-GFP. Wild-type (WT), erg2D, and erg4D strains expressing endogenously tagged Pdr12-GFP were grown to mid-log phase, centrifuged and resuspended in YPD media with or without adipic acid, and grown at 30 C for 1 h before imaging. Scale bar, 10 mm.
(D) Synergistic effects of erg4D and pdr12D on adipic acid toxicity. Wild-type (WT), erg4D (YKB4876), pdr12D (YKB4383), and erg4Dpdr12D (YKB5036) were grown to mid-log phase in YPD medium before being diluted into fresh medium to an OD 600 of 0.1. Four 10-fold serial dilutions were spotted onto YPD agar plates with or without adipic acid (60 mM) and incubated at 30 C for 48 h. The image is representative of three biological replicates. See also Figure S6. iScience Article levels were induced 12-fold after 30 min of adipic acid treatment ( Figure 5B). These data indicate that Pdr12 expression is increased during growth in adipic acid potentially to facilitate export of cytosolic adipic acid anions (adipate) in order to reduce intracellular toxicity.
As ergosterol has been reported to play a role as lipid raft in transporting newly synthesized proteins from the Golgi to the plasma membrane (Bastos et al., 2012;Mollinedo, 2012), we asked if ergosterol biosynthesis genes ERG2 and ERG4 were required for either the induction or localization of Pdr12-GFP to the plasma membrane upon adipic acid treatment. Wild-type, erg2D, and erg4D cells expressing endogenously tagged Pdr12-GFP were grown to midlog phase before being treated with 120 mM adipic acid for 1 h. We found that Pdr12-GFP in the ergosterol mutants was induced and localized to the plasma membrane similar to wild type, suggesting that Pdr12 localization to the plasma membrane is independent of ergosterol lipid rafts ( Figure 5C). Although ergosterol may not be needed to localize Pdr12 to the plasma membrane, it may still be required for Pdr12 activity. If this was the case we would hypothesize that an erg4Dpdr12D sensitivity to adipic acid would be epistatic or similar to the single mutants. If, however, these two pathways were working in parallel, the double mutant would have increased sensitivity to adipic acid compared with the single mutants. Wild-type, erg4D, pdr12D, and erg4Dpdr12D cells were tested for growth on YPD agar plates with and without a sublethal dosage of adipic acid ( Figure 5D). Although 60 mM adipic acid treatment causes slow growth of erg4D, it has limited impact on pdr12D cells. In contrast, the ability of erg4Dpdr12D cells to grow on adipic acid is dramatically slower than either erg4D or pdr12D cells. The synergistic sensitivity to adipic acid suggests that ergosterol and Pdr12 play independent roles in protecting the cells from adipic acid toxicity ( Figure 5D). Taken together, these data suggest that although Pdr12 is rapidly induced upon adipic acid treatment, its localization to the plasma membrane is not mediated by ergosterol lipid rafts. Furthermore, our genetics suggests that Erg4, and presumably the ergosterol biosynthesis pathway, and Pdr12 play non-overlapping roles in buffering adipic acid toxicity.

Adipic acid disrupts the Membrane Compartment of Can1 (MCC) and induces endocytosis
To further probe the potential roles of ergosterol in adipic acid protection, we asked if the ergosterol-rich MCC was affected by adipic acid. MCCs are membrane microdomains and occur as one of several large membrane compartments on the plasma membrane (Malinsky et al., 2010). They also serve as membrane reservoirs of specific proteins such as the proton symports Can1, Tat2, and Fur4 as well as other proteins with unknown functions including Sur7, Fmp45, and Yn1194c (Douglas et al., 2011). Can1 is an arginine transporter that interacts with ergosterol in the MCC and also serves as an MCC marker (Grossmann et al., 2008). Not only is ergosterol required to target some of these membrane proteins including Can1 and Tat2 to the MCC but also ergosterol accumulates in the MCC, which is visualized as patches when cells are stained with the filipin dye (Grossmann et al., 2007;Malínská et al., 2003). Indeed, deletion of ergosterol biosynthesis genes has been shown to impair MCC formation (Grossmann et al., 2006). To assess the effect of adipic acid on the MCC, wild-type and erg4D cells were stained with filipin before treatment with adipic acid. As expected in untreated wild-type cells, filipin staining of the plasma membrane appeared as punctate structures or MCC patches, whereas in erg4D cells punctate filipin staining is disrupted ( Figure 6A). Upon adipic acid treatment, the filipin-stained MCC patches on the plasma membrane were lost in wildtype cells ( Figure 6A), suggesting that adipic acid impacts the plasma membrane MCC. As a secondary method to assess the impact of adipic acid on the MCC, we utilized the MCC marker Can1-GFP, and as expected during growth in adipic acid, Can1-GFP is lost from the plasma membrane and is enriched in punctate internal structures ( Figures 6B and S7). Previous studies have reported that Can1-GFP undergoes endocytosis to the vacuole when it is lost from the plasma membrane (Ghaddar et al., 2014;Gournas et al., 2017). Using two vacuolar markers, CMAC and FM4-64, which stain the lumen of the vacuole and the vacuolar membrane, respectively, we determined that Can1-GFP localized to the vacuoles upon adipic acid treatment ( Figures 6B and S7). Furthermore, upon adipic acid treatment there is a significant decrease in full-length Can1-GFP protein levels and an increase in free GFP ( Figure 6C). Together this suggests that adipic acid exposure disperses Can1 from the MCC, which is subsequently internalized by endocytosis to the vacuole for degradation.
As the MCCs are sites for endocytosis on the plasma membrane, we hypothesized that adipic acid treatment induced endocytosis as a mechanism to potentially downregulate membrane transporters that import adipic acid into the cell, which results in accumulation and toxicity. To test this hypothesis, we utilized the vacuole dye FM4-64, which is also commonly used to measure endocytosis (Brach et al., 2014). We performed a time course experiment and determined that adipic acid-treated cells internalized FM4-64 ll OPEN ACCESS iScience 24, 102327, April 23, 2021 9 iScience Article Figure 6. Adipic acid disrupts the MCC and induces endocytosis (A) Adipic acid treatment disrupts ergosterol plasma membrane patches. Mid-log phase cultures of wild-type (WT, YKB1079) and erg4D (YKB4876) cells were stained with the filipin dye and treated with or without 120 mM adipic acid for 1 h, before imaging. Scale bar, 10 mm; images are representative of three biological replicates. (B) Adipic acid treatment results in Can1-GFP relocalization to the vacuole. Wild-type cells endogenously expressing Can1-GFP (YKB5097) were grown to the mid-log phase in YPD media. Cells were then stained with the vacuole lumen dye CMAC for 15 min before being treated with 120 mM adipic acid for 1 h at 30 C after which they were imaged. Scale bar, 10 mm; images are representative of three biological replicates.
(C) Can1-GFP is degraded upon adipic acid treatment. Quantitative western blot analysis were performed on whole-cell protein extracts, which had been collected by TCA extraction from Can1-GFP yeast cultures that were untreated (0 mM) or treated with adipic acid (120 mM) for 1 h. The anti-GFP blot is shown with the corresponding total protein for each lane as assessed using the Bio-Rad TGX system (left panel). The abundance of full-length Can1-GFP protein and free GFP was quantified relative to total protein and normalized to untreated levels (fold expression) (right panels). n = 3, *p < 0.05, error bars represent 1 SD. (D) Adipic acid treatment induces endocytosis. Mid-log phase cultures of the wild-type yeast (YKB1079) were stained with the FM4-64 and treated with or without adipic acid. Samples from the cultures were taken at 5, 30, 60, and 120 min after adipic acid treatment. The cells were centrifuged, resuspended in SC medium, and imaged. Scale bar, 10 mm; images are representative of three biological replicates. See also Figure S7. iScience Article faster than untreated cells ( Figure 6D). Remarkably, instead of diffused plasma membrane staining at time point 0 min, FM4-64 is already being internalized in adipic acid-treated cells, and vacuolar staining of cells by FM4-64 was visible after 15 min of incubation with adipic acid, whereas it took up to 120 min to observe clear vacuolar staining in the untreated cells. Together, these data indicate that adipic acid disrupts the membrane component of Can1 potentially by an increased rate of endocytosis.

DISCUSSION
Overcoming the challenge of adipic acid toxicity in yeast is imperative to achieve high titers of the acid in production through yeast fermentation. Therefore, in this study we sought to identify the mechanisms underlying the toxic effect of the adipic acid in yeast and identify genes that could be targeted to improve tolerance in yeast. Toward this goal, we performed a chemogenomic screen and systematically assessed the impact of each non-essential gene on adipic acid tolerance. Our screen revealed that a disruption in protein trafficking and the ergosterol biosynthesis pathway resulted in significant increase in sensitivity to adipic acid. On the contrary, deleting KGD1, a gene in the TCA cycle, resulted in improved tolerance to adipic acid.
How is KGD1 associated with enhanced tolerance to adipic acid?
In our screen, we identified one suppressor gene whose deletion resulted in tolerance to adipic acid. To the best of our knowledge, this is the first study reporting the role of the TCA cycle gene, KGD1, in increasing tolerance to adipic acid and its precursor catechol ( Figure 2). Surprisingly, only the deletion of KGD1, but not other genes in the TCA cycle, improved adipic acid tolerance (Figures 2 and S2). How is this occurring? One possibility is that deleting KGD1 results in the redirected metabolic flux toward acetyl-CoA production and this resulted in high amounts of acetate in the cell (Gruchattka and Kayser, 2015). Although we did not quantify intracellular levels of acetate in our kgd1D strain, future work should investigate if priming yeast cells with acetic acid before adipic acid exposure results in increased tolerance. This investigation will be worthwhile because acetic acid has also been shown in earlier reports to activate the plasma membrane transporter Pdr12 (Kim et al., 2019;Mira et al., 2010a). Given that Pdr12 overexpression improved tolerance to adipic acid (Pereira et al., 2019), intracellular levels of acetate in kgd1D strains may also result in Pdr12 induction and increased tolerance to adipic acid. If this was the case, however, one would anticipate that other mutants in the TCA cycle that increase cellular acetate may also confer adipic acid tolerance. However, these were not identified in the screen, potentially due to functional redundancy. Alternatively, Kgd1 may have biological functions outside of the TCA cycle. This is potentially the case because although KGD1 is a component of the mitochondrial alpha-ketoglutarate dehydrogenase complex together with KGD2 (Heublein et al., 2014), only KGD1 is associated with adipic acid tolerance. It will be interesting to further investigate why deleting KGD1 suppresses adipic acid toxicity.

How does ergosterol modulate adipic acid toxicity?
Ergosterol and other yeast sterols have been reported to modulate different stresses in yeast (Caspeta et al., 2014;Fletcher et al., 2017;Liu et al., 2017). Interestingly, in the present study we establish that cellular ergosterol also plays a role in adipic acid tolerance. Here, we show that disrupting the final steps of the ergosterol biosynthetic pathway results in severe growth defects in yeast upon adipic acid treatment (Figures 3B and S1). It is likely that ergosterol mutants lacking genes in the final steps of the pathway have variations in their sterol composition. Previously, a study reported that yeast mutants lacking enzymes coded for by these ergosterol pathway genes accumulate a mixture of sterols because Erg2, Erg3, Erg4, and Erg5, which catalyze the last steps of the pathway, are promiscuous in their choice of substrate (Abe and Hiraki, 2009). Indeed, erg2D has been shown to accumulate fecosterol (Erg2 precursor) as well as ergosta-5,8,12trienol and ergosta-8-enol, whereas erg3D accumulates a mixture of episterol, ergosta-7,22-dienol, and ergosta-7-enol (Abe and Hiraki, 2009;Johnston et al., 2020). On the other hand, erg5D accumulates ergosta-5,7,24(28)-trienol and ergosta-5,7-dienol, whereas erg4D mutant accumulates ergosta-5,7,22,24(28)tetraen-3-ol (ergosterol precursor) (Abe and Hiraki, 2009). These sterol precursors will impact the physical properties of the cell membrane in a way different from the wild-type membrane, which is made up of ergosterol as the predominant sterol. Ergosterol is a flat sterol, whereas sterol precursors such as fecosterol and episterol are bent (Caspeta et al., 2014;Shahedi et al., 2006) and will alter the cell membrane by interfering with the van der Waals's interactions within the membrane's lipid bilayer (Grossmann et al., 2007). In the context of protecting cells from adipic acid toxicity, it is possible that the membrane structure maintained by ergosterol limits the rate at which the undissociated adipic acid diffuses across the plasma membrane into the cytosol from the extracellular environment.

OPEN ACCESS
iScience 24, 102327, April 23, 2021 11 iScience Article As mentioned earlier, maintaining the right amount and composition of sterol is essential for growth in adipic acid. Changes in subcellular localization of sterol enzymes occur in cells as a way of regulating the activity of these enzymes (Leber et al., 1998). This might be the case for Erg24 and Erg4 during growth in adipic acid ( Figure 4A) particularly because Erg4 catalyzes the final step of the pathway required for ergosterol biosynthesis. Alternatively, and not mutually exclusive, membrane sterol composition may be playing a critical role in the activity and/or localization of membrane-bound proteins. For example the activity of Pdr5, an ABC transporter reported to increase yeast tolerance to certain stresses, is impaired in ergosterol biosynthesis mutants (Kodedová and Sychrová , 2015). Similarly, ergosterol has been shown to be required for Pdr12 function, which is essential for tolerance to organic acids (Xu and Li, 2020). Yeast mutants that lack genes in the ergosterol biosynthesis pathway are sensitive not only to adipic acid but also to other organic acids such as sorbic acid . Although we did not investigate whether the efflux activity of Pdr12 is also significantly impaired in ergosterol mutants, we showed that deleting Pdr12 in an ergosterol mutant resulted in a synergistic effect on growth in adipic acid. The erg4Dpdr12D mutant was more sensitive to adipic acid than the erg4D or pdr12D alone ( Figure 5D). This observation suggested that ergosterol plays other roles in modulating adipic acid toxicity beyond a potential role in regulating Pdr12. Regarding its role in the localization of membrane-bound proteins, ergosterol transports and targets membrane proteins such as Can1 and Tat2 to the MCC via ergosterol lipid rafts (Grossmann et al., 2008). Furthermore, ergosterol along with several proteins forms the MCC, which are sites of endocytosis on the plasma membrane, and defects in MCC formation results in impaired endocytosis (Heese-Peck et al., 2002). From our data, adipic acid impacts endocytosis by disrupting MCC formation, which was evidenced in the rapid internalization of Can1-GFP and FM4-64 ( Figure 6D).
Why is protein trafficking and vacuolar transport essential for yeast growth in adipic acid?
Nearly half of the genes in the adipic acid chemogenomic profile are associated with protein trafficking and vacuolar transport, suggesting that protein degradation and recycling in the vacuole is an essential biological process required for survival during yeast growth in adipic acid. Interestingly, most of these genes code for proteins that are components of the complexes that form the Endosomal Sorting Complexes Required for Transport (ESCRT) machinery. The ESCRT pathway is used by cells to degrade specific proteins on the plasma membrane by endocytosis (Wenzel et al., 2018). Intriguingly, our study shows that adipic acid impacts and disrupts the MCC ( Figure 6A), which serves as a site for endocytosis and also provides a protective surrounding for several plasma membrane proteins and transporters including Can1 permease. A previous study has shown that when the protective area offered by the MCC is disrupted, proteins that reside in this compartment are released and become susceptible to cellular internalization and endocytosis (Grossmann et al., 2007(Grossmann et al., , 2008. This explains why when adipic acid disrupts the MCC, Can1 rapidly changes localization to the vacuole for degradation ( Figures 6B and 6C). This explanation was independently strengthened by the fact that adipic acid induced endocytosis ( Figure 6D).
Endocytosis is essential for degrading and recycling of plasma membrane proteins. Specific membrane proteins are ubiquitinylated, internalized by endocytosis, and transported to the endosomes via the ESCRT pathway for vacuolar degradation (Feyder et al., 2015). Although organic acids are generally known to enter cells by passive diffusion, it has been shown that some plasma membrane transporters such as Jen1 and Acy1 are able to transport lactate, acetate, and other organic acids into the cell (Casal et al., 2008;McDermott et al., 2010). We speculate that endocytosis is induced in adipic acid-treated cells as a way of degrading non-specific membrane transporters that import extracellular adipic acid into the cell in order to prevent intracellular accumulation of the acid. It will be interesting to screen and identify non-specific transporters that import adipic acid into the cell and determine how adipic acid treatment impacts their degradation by endocytosis. Furthermore, as endocytosis is induced in adipic acid-treated cells as a survival mechanism, it is imperative that all the proteins involved in this process such as the ESCRT pathway proteins are maintained. Hence, when any of these key proteins are disrupted by deleting the genes that encode them, the cells become hypersensitive to adipic acid as shown in the adipic acid chemogenomic profile in this study.

Conclusion
In conclusion, our study provides an outlook into the basic cellular mechanisms underlying adipic acid toxicity in yeast and provides potential genetic candidates that can be engineered for tolerance to boost adipic acid production in yeast. The results presented in this study highlight an induction of endocytosis iScience Article in cells treated with adipic acid. It is therefore essential to maintain ergosterol biosynthesis because ergosterol is required to maintain the integrity of the plasma membrane as well as the MCC needed for endocytosis. Furthermore, KGD1 deletion suppresses the toxicity of adipic acid and its toxic catechol precursor, making this gene an important target for engineering robust adipic acid production strains.

Limitations of the study
The present study demonstrated an important role for the last four steps of the ergosterol biosynthesis pathway in protecting yeast cells from the toxic effects of adipic acid. Further studies will be required to determine if modulating the ergosterol pathway, such as overexpression, can improve growth upon adipic acid exposure. In addition, as this work was conducted in the laboratory yeast strain background S288c further studies are required to determine if the findings presented here are transferrable to industrial yeast strains.

Resource availability Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kristin Baetz (kbaetz@uottawa.ca).

Materials availability
This study did not generate new unique reagents. Yeast strains generated in this study that are not available commercially are available from the lead contact without restriction with the requestor paying for shipping costs.

Data and code availability
The published article includes all datasets generated or analyzed during this study.

METHODS
All methods can be found in the accompanying transparent methods supplemental file.