The Rpd3-Complex Regulates Expression of Multiple Cell Surface Recycling Factors in Yeast

Intracellular trafficking pathways control residency and bioactivity of integral membrane proteins at the cell surface. Upon internalisation, surface cargo proteins can be delivered back to the plasma membrane via endosomal recycling pathways. Recycling is thought to be controlled at the metabolic and transcriptional level, but such mechanisms are not fully understood. In yeast, recycling of surface proteins can be triggered by cargo deubiquitination and a series of molecular factors have been implicated in this trafficking. In this study, we follow up on the observation that many subunits of the Rpd3 lysine deacetylase complex are required for recycling. We validate ten Rpd3-complex subunits in recycling using two distinct assays and developed tools to quantify both. Fluorescently labelled Rpd3 localises to the nucleus and complements recycling defects, which we hypothesised were mediated by modulated expression of Rpd3 target gene(s). Bioinformatics implicated 32 candidates that function downstream of Rpd3, which were over-expressed and assessed for capacity to suppress recycling defects of rpd3∆ cells. This effort yielded three hits: Sit4, Dit1 and Ldb7, which were validated with a lipid dye recycling assay. Additionally, the essential phosphatidylinositol-4-kinase Pik1 was shown to have a role in recycling. We propose recycling is governed by Rpd3 at the transcriptional level via multiple downstream target genes.


Introduction
Most integral membrane proteins expressed in eukaryotic cells are inserted into the endoplasmic reticulum (ER) via different mechanisms [1]. Many of these perform diverse roles at the plasma membrane (PM), such as acting as ion channels, nutrient transporters, and different classes of receptors [2,3]. Surface proteins are actively transported from the ER through the secretory pathway to the cell surface [4]. Mechanisms of surface protein regulation have been elucidated using the budding yeast Saccharomyces cerevisiae, where hundreds of proteins are organised in distinct spatial arrangements [5]. The lateral movement of proteins between regions of the PM correlates with their biological activity [6]. For example, inactive nutrient transporters localised to eisosome subdomains adopt an active conformation for nutrient uptake upon migration to other regions of the PM in response to a substrate [7,8]. This altered PM localisation of active transporters supports their internalisation and endocytosis [9], a process which is controlled metabolically, with stress conditions altering eisosomal capacity to harbour nutrient transporters [10,11].
Surface proteins are internalised to the endosomal system, a network of intracellular compartments that organise and traffic protein and lipid material to other intracellular

The Rpd3 Complex Is Required for Efficient Surface Protein Recycling
Endosomal recycling can be tracked using the Ste3-GFP-DUb recycling reporter [18], which recycles efficiently in wild-type cells but is retained in intracellular endosomes in recycling mutants such as rcy1∆ cells (Figure 1A), demonstrated by 3D confocal projections of the recycling reporter in wild-type cells and recycling defective rcy1∆ mutants (Movie S1). Plasmid expression of Ste3-GFP-DUb in 4985 haploid deletion mutants revealed 89 validated mutants that are defective in cell surface recycling. Among this list of potential recycling factors was rpd3∆ cells, which lack a histone modifying regulator of gene expression [33], alongside nine other mutants of the Rpd3-complex [44] ( Figure 1B). To confirm these results, we stably integrated Ste3-GFP-DUb into wild-type cells and all 10 previous identified mutants of the Rpd3 complex, revealing all mutants had some degree of recycling defect, with intracellular accumulation of reporter similar to that observed in rcy1∆ cells ( Figure 1C). To quantify differences between the null strains lacking components of the Rpd3-complex, we optimised the segmentation of cells using phase contrast (PC) and digital interference contrast (DIC). Although PC introduced an obvious border to define cells for segmentation in brightfield micrographs, we noted a significantly poorer fluorescence signal (Figure 2A).
We therefore refined parameters to define cells from DIC images and used this protocol to estimate the background autofluorescence in the green channel capturing Ste3-GFP-DUb expressing yeast cells. For this, cells expressing Ste3-GFP-DUb were mixed with a separate culture of cells expressing Gpa2-mCherry and imaged ( Figure 2B). Gpa2 is a Gαsubunit that regulates cAMP production that exclusively localises to the periphery via lipid modifications [22,45]. Segmentation of cells expressing Gpa2-mCherry allowed cellular autofluorescence to be measured in optical conditions optimised for Ste3-GFP-DUb acquisitions ( Figure 2C). Having optimised segmentation and normalisation parameters, we then used a morphological erosion function [46] to measure the levels of cell surface signal as a percentage of total fluorescence. As expected, Ste3-GFP-DUb and Gpa2-mCherry, which both primarily localise to the PM at steady state, had similarly high levels of PM localisation ( Figure 2D). Applying this analysis to all mutants of the Rpd3-complex expressing Ste3-GFP-DUb revealed every mutant had defective recycling compared to wild-type cells ( Figure 2E). We found the most defective mutant was hos2∆, lacking the Hos2 deacetylase, which was as defective as the prototypical recycling mutant strain rcy1∆. The least defective mutant was ume1∆, lacking the Ume1 co-repressor, which still exhibited significantly impaired recycling compared to wild-type cells.
To validate the role of the Rpd3 complex in surface recycling, we have previously employed an assay that takes advantage of recycling mutants exhibiting defective surface localisation of the tryptophan permease Tat2 ( Figure 3A), which is required for growth in limited tryptophan media [47]. This assay requires tryptophan auxotroph strains, so we tested Rpd3-complex mutants in the SEY6210 background that harbour a trp1-∆901 mutation for growth capacity in media of replete (40 mg/L) versus low tryptophan concentrations (5 and 2.5 mg/L). We have recently documented a quantitative analysis method for such growth assays across a large spectrum of serial dilutions [48], which we used to quantify growth defects attributed to defective Tat2 recycling in Rpd3-complex mutants ( Figure 3B,C). We confirmed that most mutants had a low tryptophan-dependent phenotype. Importantly, only pho23∆ cells, which lack the Pho23 regulator of Rpd3 activity [49], exhibited wild-type like growth in both low (5 and 2.5 mg/L) tryptophan media, and ume1∆ cells in 2.5 mg/L tryptophan had growth that was not significantly different to wild-type. Convincingly, pho23∆ and ume1∆ mutants were the two least defective mutants from our quantitative analysis of Ste3-GFP-DUb recycling in a completely different strain background. This indirect assay further suggests that Tat2 recycling is perturbed in all these mutants lacking Rpd3-subunits. We note that expression of TAT2 in rpd3∆ cells is very similar to wild-type cells [50,51], but even when Tat2-3xHA was expressed from an endogenous promoter, levels were significantly reduced in rpd3∆ cells ( Figure 3D,E). We assume the reduced recycling of Tat2 that impairs tryptophan uptake results in increased trafficking of Tat2 from endosomes to the vacuole for degradation.   shown overlaid on each fluorescence image. A second ROI that excludes the plasma membrane signal was created by morphological erosion. (B,C) Wild-type cells expressing Gpa2-mCherry from the CUP1 promoter, induced by the addition of 100 µM copper chloride to the media, and wild-type cells expressing Ste3-GFP-DUb were mixed at a 1:1 ratio and imaged. mCherry-expressing cells were used to identify and segment the wild-type cells lacking GFP signal, allowing cellular background fluorescence in the green channel, observed at increased intensity (right), to be measured. (D) The surface levels of Ste3-GFP-DUb and Gpa2-mCherry were both calculated as a percentage. (E) The percentage of Ste3-GFP-DUb signal at the plasma membrane in wild-type cells (grey, n = 104) was compared with rcy1∆ mutants (green, n = 111) and various mutants (magenta) of the Rpd3 complex: hos2∆ (n = 47), cti6∆ (n = 73), sin3∆ (n = 114), rpd3∆ (n = 57), stb2∆ (n = 61), pho23∆ (n = 77), and ume1∆ (n = 36). Student's t-test comparisons between wild-type cells and each mutant are shown with asterisks (*) used to indicate significant difference (p < 0.0001). Scale bar, 5 µm.
We therefore refined parameters to define cells from DIC images and used this protocol to estimate the background autofluorescence in the green channel capturing Ste3- shown overlaid on each fluorescence image. A second ROI that excludes the plasma membrane signal was created by morphological erosion. (B,C) Wild-type cells expressing Gpa2-mCherry from the CUP1 promoter, induced by the addition of 100 µM copper chloride to the media, and wild-type cells expressing Ste3-GFP-DUb were mixed at a 1:1 ratio and imaged. mCherry-expressing cells were used to identify and segment the wild-type cells lacking GFP signal, allowing cellular background fluorescence in the green channel, observed at increased intensity (right), to be measured. (D) The surface levels of Ste3-GFP-DUb and Gpa2-mCherry were both calculated as a percentage. (E) The percentage of Ste3-GFP-DUb signal at the plasma membrane in wild-type cells (grey, n = 104) was compared with rcy1∆ mutants (green, n = 111) and various mutants (magenta) of the Rpd3 complex: hos2∆ (n = 47), cti6∆ (n = 73), sin3∆ (n = 114), rpd3∆ (n = 57), stb2∆ (n = 61), pho23∆ (n = 77), and ume1∆ (n = 36). Student's t-test comparisons between wild-type cells and each mutant are shown with asterisks (*) used to indicate significant difference (p < 0.0001). Scale bar, 5 µm. Yeast were grown to mid-log phase and spotted out on media of replete (40 µg/mL) and limited, either 5 µg/mL in (B) and 2.5 µg/mL in (C) tryptophan concentration. Growth was measured across multiple serial dilutions and calculated as a ratio compared to wild-type cells from the same plate. Asterisks (*) used to indicate significant difference (p < 0.03) from t-test comparisons. (D) Cells transformed with a Tat2-3xHA plasmid containing a copper-inducible CUP1 promoter were grown in media containing 50 µM copper chloride to mid-log phase before lysates were generated for immunoblot analysis using α-HA and α-Rsp5 antibodies. (E) Densitometry was used to measure the signal intensity of Tat2-HA in different clones and strains from (D), normalized to loading control. In tryptophan auxotroph cells grown on media containing low tryptophan concentrations, Tat2 uptake is required for efficient growth of wild-type cells (left) and Tat2 recycling defects inhibit growth (left). (B,C) Yeast were grown to mid-log phase and spotted out on media of replete (40 µg/mL) and limited, either 5 µg/mL in (B) and 2.5 µg/mL in (C) tryptophan concentration. Growth was measured across multiple serial dilutions and calculated as a ratio compared to wild-type cells from the same plate. Asterisks (*) used to indicate significant difference (p < 0.03) from t-test comparisons. (D) Cells transformed with a Tat2-3xHA plasmid containing a copper-inducible CUP1 promoter were grown in media containing 50 µM copper chloride to mid-log phase before lysates were generated for immunoblot analysis using α-HA and α-Rsp5 antibodies. (E) Densitometry was used to measure the signal intensity of Tat2-HA in different clones and strains from (D), normalized to loading control.

Hypothesis for the Role of the Rpd3-Complex in Recycling
Having confirmed and quantified the requirement of the Rpd3 complex in surface recycling of diverse cargoes, we set out to reveal the mechanisms regulating this trafficking pathway. As previously documented, Rpd3 primarily localises to the nucleus [34]. As expected, Rpd3 with a C-terminal mCherry tag (Rpd3-mCherry) localised to the nucleus, labelled with Hoechst, but also to the cytoplasm ( Figure 4A). Similar results were observed for an N-terminally tagged fusion of mCherry-Rpd3 ( Figure 4B). The Ste3-GFP-DUb recycling defect of rpd3∆ cells was successfully rescued by the expression of tagged Rpd3, showing these fusion proteins are functional ( Figure 4C). The nuclear localisation of the KDAC Rpd3 allows regulation of transcription via the post-translational modification of chromatin [31]. Rpd3 forms both small and large complexes with Sin3 and other subunits that contribute to its regulation [44,52], with various additional studies further documenting protein-protein interactions between Rpd3-complex subunits [53][54][55][56][57], which we highlight in a physical interaction network ( Figure 4D). Importantly, many of these factors (10 out of 14) were independently identified from a blind genetic screen for recycling machinery [19] and subsequently validated and quantified (Figures 1-3). We hypothesised that the Rpd3complex regulates the expression of either specific recycling factor(s) identified from the Ste3-GFP-DUb localisation screen or regulates the expression of an unknown essential gene not represented in the library of viable haploid deletions used for this screen.

Downstream Rpd3 Targets Regulate Recycling
As recycling defects were phenocopied across most mutants lacking subunits of the Rpd3-complex, we reasoned that gene expression differences of potential target genes would be shared across mutants. Therefore, we assembled gene expression profiles [51] for the 89 validated recycling factors [19], depicted as a heatmap ( Figure 5A) but averaged the changes in gene expression across all mutant conditions. This allowed us to identify 22 mutants that are significantly over-expressed in mutants of the Rpd3-complex. However, over-expressing these factors in wild-type cells co-expressing Ste3-GFP-DUb did not induce any defects (Supplemental Figure S1). We also identified 24 genes with significantly reduced expression across mutants. These represented our candidates for a complementation screen, as we hypothesised their repression via Rpd3-complex results in recycling defects. Therefore, reintroducing high levels of these factors might supress these recycling defects. We also included the next 10 genes that were decreased to a smaller degree, to account for gene regulation that was repressed significantly in only certain mutants (for example to account for technical errors during transcriptomic analyses). Each of these genes were over-expressed from a plasmid library [58] in rpd3∆ cells stably expressing Ste3-GFP-DUb. We were unable to test complementation of GPA1 and HDA1, as these clones repeatedly did not yield any transformations, potentially as the combinations of these over-expressors with rpd3∆ are not viable. We assessed reporter localisation in each over-expression condition from three independent transformants and used a qualitative scoring system to document results ( Figure 5B). This screen revealed three factors: Sit4, Dit1 and Ldb7, that complement the recycling defect of rpd3∆ cells, which were all confirmed by further imaging experiments ( Figure 5C) and quantification using the analysis pipeline discussed above ( Figure 5D). We included images and analysis of Prm8 as a control, as PRM8 was the most repressed gene that failed to complement recycling upon over-expression. Although Sit4 has been implicated in trafficking [59,60] and the others have not, we also validated complementation of sin3∆ and dep1∆ recycling defects with DIT1 and LDB7 over-expression (Supplemental Figure S2). Rpd3-mCherry were grown to mid-log phase, washed twice in fresh media and then incubated for 30 min with water containing 8 µM Hoechst, followed by fluorescence microscopy. (B) Yeast cells transformed with plasmids expressing Rpd3 from the CUP1 promoter with either a C-terminal (left) or N-terminal (right) mCherry tag were grown to mid-log phase in media containing 50 µM copper chloride prior to confocal microscopy. (C) Mutant rpd3∆ cells expressing an endogenously expressed version of Ste3-GFP-DUb were transformed with a vector control (upper) and a plasmid encoding Rpd3-mCherry under control of the CUP1 promoter (lower). Transformants were grown to mid-log phase prior to confocal microscopy. (D) A protein association network based only on experimental evidence of physical interactions (confidence = 0.400) was generated for the Large Rpd3 complex using STRING v11.5. Entries are coloured based on a k-means clustering algorithm for 3 clusters (red, green, blue) but also outlines indicate whether mutants of these proteins were identified (black) or not (blue) from the Ste3-GFP-DUb localisation screen. Scale bar, 5 µm. Rpd3-mCherry were grown to mid-log phase, washed twice in fresh media and then incubated for 30 min with water containing 8 µM Hoechst, followed by fluorescence microscopy. (B) Yeast cells transformed with plasmids expressing Rpd3 from the CUP1 promoter with either a C-terminal (left) or N-terminal (right) mCherry tag were grown to mid-log phase in media containing 50 µM copper chloride prior to confocal microscopy. (C) Mutant rpd3∆ cells expressing an endogenously expressed version of Ste3-GFP-DUb were transformed with a vector control (upper) and a plasmid encoding Rpd3-mCherry under control of the CUP1 promoter (lower). Transformants were grown to mid-log phase prior to confocal microscopy. (D) A protein association network based only on experimental evidence of physical interactions (confidence = 0.400) was generated for the Large Rpd3 complex using STRING v11.5. Entries are coloured based on a k-means clustering algorithm for 3 clusters (red, green, blue) but also outlines indicate whether mutants of these proteins were identified (black) or not (blue) from the Ste3-GFP-DUb localisation screen. Scale bar, 5 µm.  ume1∆, hos2∆, cti6∆, pho23∆, sap30∆, dep1∆, sin3∆, rpd3∆) as a heat map (lower right). The successful deletion of each gene is shown to reduce expression of each individual subunit (lower left). (B) Each of the listed genes were chosen for over-expression in rpd3∆ cells stably integrated with Ste3-GFP-DUb. Single colony transformants of each were grown to mid-log phase and imaged by confocal microscopy. Transformants that rescue Ste3-GFP-DUb recycling were first scored qualitatively, with modest (light green) and substantial (dark green) levels of potential complementation of recycling indicated. (C) Over-expression candidates identified from the screen described in (B) were grown to mid-log phase and prepared for confocal imaging. (D) Quantification of % PM Ste3-GFP-DUb in each of the indicated cellular conditions, with asterisks (*) used to indicate significant difference of p < 0.0001 from t-test comparisons. Scale bar, 5 µm. We first confirmed the role of Sit4, Dit1 and Ldb7 in recycling Ste3-GFP-DUb (Supplemental Figure S3). In order to validate these complementation factors, we employed a distinct recycling assay based on the lipid dye FM4-64, which can be loaded to endosomes for brief labelling periods, followed by tracking efflux via recycling that triggers dye quenching [20]. We found that the rate of efflux from recycling mutants is reduced ( Figure 6A). Efflux can be tracked through a kinetic assay using flow cytometry, measuring fluorescence from~1500 cells per second and the fluorescence of each cell is presented as a percentage of the initial fluorescence, calculated as the average from all events in the first 10 s ( Figure 6B). This assay was used to show that plasmid over-expression of Sit4, Dit1 and Ldb7 all increase the rate of FM4-64 recycling observed in rpd3∆ cells ( Figure 6C-F). A wild-type positive control, and rpd3∆ cells transformed with an empty vector as a negative control, were performed at the same time and the profiles of these efflux measure ments overlaid to each complementation profile.

Rpd3 Regulates the Phosphatidylinositol 4-Kinase Pik1
The Rpd3 complex might also regulate expression of essential genes that are involved in membrane trafficking of surface proteins. To explore this possibility, we used recently optimised bioinformatics approaches [48] to assemble gene expression profiles of only essential genes that were not tested in the original recycling reporter screen ( Figure 7A). We compared essential gene profiles of seven different null strains representing Rpd3-complex members [51], many of which share large regions of expression patterns with hos2∆ and ume1∆ cells being most distinct ( Figure 7B). As the recycling phenotypes are shared across the various mutants, we averaged the changes in expression across mutants to identify those with most significantly altered expression. Gene ontology analyses of the most repressed 43 genes (log2 fold change <5.0) was performed and showed enrichment for processes including autophagy, phosphorylation and lipid regulation ( Figure 7C), with a large amount of annotation overlap of the most enriched ( Figure 7D). We considered PIK1, an essential phosphatidylinositol-4-kinase (PI4K) that regulates trafficking from both the Golgi and endosomes [61][62][63][64], would be a likely Rpd3 target gene with the potential to regulate trafficking of surface membrane proteins. Indeed, the levels of PIK1 are substantially decreased in mutants of the Rpd3-complex, even when viewed with all essential and non-essential gene profiles ( Figure 7E). To test the hypothesis that Pik1 is required for efficient recycling, we again employed the FM4-64 efflux assay ( Figure 6A). For this, wild-type cells and temperature-sensitive mutants were loaded with FM4-64 and the rate of efflux was measured over time. There was a significant decrease in recycling efficiency in cells expressing either a mutant allele of pik1 (pik1-83 and pik1-139), even when the cytometry experiments were performed in media that was not at a restrictive temperature ( Figure 7F).
It has previously been proposed that the unfolded protein response (UPR), which is elevated in rpd3∆ cells, results in surface proteins like the uracil permease Fur4 to be retained in the ER and degraded [41]. This conclusion was based on dramatically reduced cellular levels of fluorescently tagged Fur4, not increased levels of GFP retained following vacuolar degradation. However, Fur4 tagged with mNeonGreen (mNG) expressed in wild-type and rpd3∆ cells localises at the PM and inside the vacuole, with no indication of ER retention ( Figure 8A). Furthermore, we observed no difference in overall levels of Fur4-mNG between wild-type cells and rpd3∆ mutants ( Figure 8B). For this reason, we propose the regulation of Rpd3 on cell surface proteins is not indirect via the UPR, but through regulation of factors required for recycling. We conclude that both non-essential and essential gene targets of the Rpd3-complex have the capacity to regulate cell surface recycling governed at the transcriptional level ( Figure 8C).

Rpd3 Regulates the Phosphatidylinositol 4-Kinase Pik1
The Rpd3 complex might also regulate expression of essential genes that are involved in membrane trafficking of surface proteins. To explore this possibility, we used recently

Discussion
The lysine deacetylase Rpd3, alongside many of its physical interactors, is known to have massive effects on gene expression in yeast [51,[65][66][67]. The identification of all 10 members of the complex from a blind screen [19] strongly implicates the Rpd3 complex as a regulator of recycling in yeast. In this study, we showed that these 10 subunits are required for recycling, but to different degrees. We stably integrated Ste3-GFP-DUb to give consistent phenotypes over plasmid-borne expression initially used to screen for factors. This allowed a quantification approach to specifically measure how much GFP signal was found efficiently recycled to the PM versus signal retained in endosomes ( Figure 2). Importantly, these results were compared to the quantitation of a growth defect indirectly associated with Tat2 recycling (Figure 3). For example, hos2∆ mutants were among the most defective mutants in both quantified assays, and both ume1∆ and pho23∆ were the least defective. We note that unlike Ste3-GFP-DUb localisation, the tryptophan-uptake assay was not sensitive enough to identify defects in Tat2 recycling in pho23∆ mutants. Although Rpd3 is associated with global deacetylation events [31], Pho23 is more limited to a specific subset of loci [49,68], so it may be that Pho23 does not regulate all genes associated with efficient recycling. We show the DNA-binding protein Ume6 is required for recycling, with ume6∆ mutants one of the most defective Tat2 recycling mutants. The genetic screen did not identify several Rpd3 subunits, such as the DNA-binding protein Ash1 (Figure 4), which exhibits gene specific regulation with Ume6 [69]. Therefore, many, but not all, Rpd3-subunits are required for efficient transcriptional control of the recycling pathway.
To explain these results, a phenotypic complementation screen was performed by over-expressing downstream recycling genes repressed in rpd3∆ mutants. We predicted this over-expression strategy, using 2 µ based plasmids retained with 100 s of copies per cell [70], would overwhelm repression mechanisms in rpd3∆ cells to reveal bona fide target recycling genes. A total of 32 candidates were tested in this screen, with only two additional candidates, HDA1 and GPA1, failing to yield viable transformants after various attempts and optimisations. Higher levels of Hda1, which is histone deacetylase related to Rpd3 with shared molecular activity [31], might induce lethality in rpd3∆ mutants. This screen revealed three validated hits that rescue recycling of rpd3∆ mutants: the protein phosphatase Sit4, the transcriptional regulator Ldb7 and the sporulation factor Dit1 (Figures 5 and 6). The Sit4 phosphatase is a strong candidate for regulating recycling as it has been previously shown to modify machinery in secretory [60] and endocytic [59] trafficking pathways. Ldb7 is itself a transcriptional regulator [71], in the family of low-dye-binding mutants associated with Golgi function, stress response and cell wall organisation [72], any of which might indirectly impinge recycling. Finally, Dit1 regulates the formation of spore walls following developmental expression of starved diploid cells [73,74]. DIT1 is repressed in haploids via the Ssn6-Tup1 repressor [75] and increased greatly during sporulation [76]. However, many independent studies report the haploid expression of DIT1, that can be further repressed upon increased temperature [77], deletion of RPD3 [50] or addition of the Rpd3 inhibitor trichostatin A [78], potentially pointing to a distinct function. Finally, the implication of the Pik1 in recycling (Figure 7) is easily rationalised as it is known to modify lipids required for proper Golgi and endosomal trafficking [61], so fine tuning of these various pathways that control surface proteins could be mediated via the Rpd3-complex.
Our observations are consistent with other reports in the literature relating to surface protein regulation and the Rpd3-complex. For example, the Trk2 surface potassium channel has been proposed to be negative regulated in cell mutants of rpd3 [79]. Similarly, although rpd3∆ cells have increased expression of the major acid phosphatase PHO5 [80], phosphate uptake via the surface localised H + /PO 4 3− symporter Pho84 is defective in rpd3∆ cells, with accumulation of Pho84-GFP in endosome-like compartments in cells lacking RPD3 [81]. Encouragingly, both our bioinformatic and functional observations on recycling align with a study on how yeast cells respond to exposure to the anti-malaria drug artemisinin, which found that both rpd3∆ and sit4∆ cells were hypersensitive to the drug and suggested this is due to premature degradation of surface proteins [41]. Beyond this, we have previously shown that the developmentally regulated expression of the Cos proteins, which drive ubiquitin-mediated vacuolar degradation of surface proteins in trans [82], is ablated by deletion of either RPD3 or SIN3 [83]. This demonstrates that complex and overlapping modes of transcriptional regulation control the membrane trafficking routes used by cell surface membrane proteins.
The experimental validation of these candidates demonstrates the complexity of surface protein recycling in yeast. As Rpd3 and some of the downstream recycling gene targets discussed are highly conserved throughout evolution, this regulatory control could be maintained in other eukaryotic systems. Indeed, Rpd3 has orthologues expressed in various other eukaryotic systems, and its roles could be understood in terms of regulating surface membrane proteins. For example, in Drosophila melanogaster, Rpd3 is required for cells to respond appropriately to nutrient starvation [84]. Regarding the discoveries in yeast related to recycling and its metabolic or transcriptional control, we recommend using more than one assay, such as Ste3-GFP-DUb localisation (Figure 1), Tat2-mediated tryptophan update (Figure 3), or efflux of internalised FM4-64 ( Figure 6), to fully validate any defect. Beyond this, we promote the use of bioinformatics to identify downstream targets of the transcriptional regulators for experimental testing, as discussed in this study. Future work will be aimed at deciphering the individual roles of these new candidates in the recycling pathways and understanding any functional overlap.

Reagents
Yeast strains and plasmids used are included as Supplemental Tables S1 and S2, respectively.

Yeast Strains and Culture Conditions
Yeast strains used in this study are listed in Table S1 and were grown in either yeast extract peptone dextrose (YPD) media, for example when making competent cells, or synthetic complete (SC) minimal drop-out media lacking appropriate bases/amino acids, when selection of plasmids or integrations was necessary. Competent yeast stocks were prepared in Li-TE sorbitol buffer (100 mM lithium acetate, 10 mM Tris.HCl pH 7.5, 1.2 M sorbitol, 1 mM EDTA, 200 µM calcium chloride) and plasmids incubated for 40 min at 30 • C followed by heat shock at 42 • C for 20 min and plating on solid selective media. Yeast cultures were prepared by inoculation from a clonal yeast patch and grown overnight at 30 • C in 5 mL two-fold serial dilutions to ensure cells used in experimental procedures were at mid log (OD 600 =~1.0). Expression of proteins from the CUP1 promoter was induced by addition of 50 µM Copper Chloride (CuCl 2 ) to the media for at least 1 h prior to experiments. Nuclei of yeast cells were labelled with fluorescent DNA stain by first growing to mid-log phase, washing in fresh SC media, prior to addition of 8 µM Hoechst-33342 (Invitrogen TM , Eugene, OR, USA) for 30 min.

DNA Manipulations
Yeast expression plasmids were purified from~5ml saturated cultures using a Wizard ® DNA Purification System (Promega, Madison, WI, USA) and were transformed into competent yeast. The Gibson assembly principle [85] of incubating homologous PCR products with Taq ligase (NEB, Ipswich, MA, USA), T5 exonuclease (NEB) and Phusion polymerase (NEB) for 1 h at 50 • C, followed by plating on selective 2YT media, was used to create different fluorescently labelled Rpd3 expression constructs, which were confirmed by Sanger sequencing. Stable integrations of Ste3-GFP-DUb under control of the STE3 promoter were performed by linearising pCM850 with NsiI followed by ethanol precipitation and transformation into the various parental yeast strains. For strains with loxP flanked integrations, cassettes were excised using a modified TEF1-Cre expression system [86].

Fluorescence Microscopy
Yeast cells were grown to mid-log phase and prepared for confocal microscopy by centrifugation to concentrate samples and resuspension in water prior to storage on ice prior to confocal microscopy. Imaging was performed from live cells using a laser scanning confocal microscope (Zeiss LSM 780, Oberkochen, Germany) with a 63× Differential Interference Contrast (DIC) or 63× Phase-Contrast (PC) oil-immersion objectives (Objective Plan-Apochromat 63×/1.4 Oil, Numerical Aperture 1.4, Zeiss). Fluorescence microscopy images were captured via Zen Black (Zeiss) software and modified for contrast, colour and merge in ImageJ (version 2.0.0).

Bioinformatics and Statistics
For non-essential ( Figure 5) and essential (Figure 7) recycling gene candidates, gene expression profiles were averaged across all mutants, since all mutants phenocopy one another with regards to recycling defects, and prioritised based on this average. Heat maps were generated to show distribution of individual values across mutants. Microarray data documenting expression changes (log fold and p values) were read into RStudio (version 1.2.5033 RStudio Team, 2020) then processed using the dplyr (v1.0.7) [88] and tidyverse (v1.3.0) [89] packages to include only data for indicated deletion strains of the Rpd3-complex. The data were further sub-setted to include genes deemed essential for viability (1290 ORFs represented) acquired from the Saccharomyces Genome Database. Hierarchical clustering was visualised in the form of a heatmap using base R. Gene Ontology enrichments were performed using GO Term Finder (version 0.86) via YeastMine [90,91].
Physical interaction maps were generated using STRING pathway (version 11.5) analysis software [92]. Statistical analyses were performed using Graphpad (Prism, version 9.0.2, San Diego, CA, USA).